Appendix D: Datasets Description

Introduction

This appendix provides a comprehensive reference guide to all datasets used throughout the 15-week linear models course. Each dataset is carefully designed to illustrate specific statistical concepts while maintaining biological realism relevant to animal breeding and genetics.

Purpose and Scope

This reference is designed to:

Document all datasets: Complete descriptions of 48 datasets spanning 5 livestock species
Provide context: Explain the pedagogical purpose and biological relevance of each dataset
Enable quick lookup: Multiple access paths (by week, species, size, or alphabetically)
Support learning: Variable descriptions and special features clearly noted

How to Use This Appendix

Week-by-week sections: Find datasets organized chronologically (Weeks 1-15)
Species Index (Species Index): Browse datasets by livestock type
Size Guide (Dataset Size Guide): Find hand-calculable vs. realistic datasets
Quick Reference (Quick Reference: Alphabetical Listing): Alphabetical listing of all datasets

Relationship to Course Structure

Datasets are organized to match the 15-week course progression: - Weeks 1-3: Small examples for matrix operations and design matrices - Weeks 4-6: Regression datasets (simple and multiple) - Weeks 7-10: ANOVA family (one-way, contrasts, two-way, ANCOVA) - Week 11: Diagnostic datasets with special features (outliers, influential points) - Weeks 12-14: Advanced topics (unbalanced, constraints, special cases) - Week 15: Comprehensive capstone integration

Summary Statistics

Overview

Metric	Value
Total datasets	48
Total observations	2,011
Size range	4-177 observations
Species covered	5 (Dairy, Beef, Swine, Poultry, Sheep)
Weeks with datasets	14 (Week 2 is pure theory)

Distribution by Week

Week	# Datasets	Total Obs	Statistical Focus
1	1	30	Matrix notation, sample mean
2	0	0	Linear algebra theory
3	2	26	Design matrix construction
4	4	59	Simple linear regression
5	1	40	Least squares theory
6	1	50	Multiple regression
7	4	123	One-way ANOVA
8	2	62	Contrasts, estimable functions
9	3	74	Two-way ANOVA, interactions
10	4	119	ANCOVA, adjusted means
11	8	425	Model diagnostics
12	4	300	Non-full rank, unbalanced
13	5	422	Special topics, constraints
14	8	300	Polynomial, WLS, mixed model preview
15	1	~120 (planned)	Capstone integration

Distribution by Species

Species	# Datasets	Avg Size	Size Range	Primary Traits
Dairy Cattle	13	52	12-100	Milk yield, lactation curves, SCC
Beef Cattle	9	56	9-177	ADG, marbling, carcass traits
Swine	8	26	6-162	Litter size, ADG, growth
Poultry	12	41	4-89	Egg production, body weight, FCR
Sheep	6	32	5-72	Weaning weight, fleece, growth

Distribution by Size Category

Category	Size Range	# Datasets	Purpose	Example Weeks
Hand-calculable	4-12	15	Manual computation practice	1, 3, 4, 7, 8, 9, 12
Moderate	20-50	25	Realistic applications	4, 5, 6, 7, 8, 10, 13, 14
Large	60-177	8	Complex structures, diagnostics	11, 12, 13, 14

File Conventions

Directory Structure

All datasets are stored in week-specific subdirectories:

WeekXX_TopicName/data/dataset_name.csv

Example: - Week 4 datasets: Week04_SimpleRegression/data/ - Week 7 datasets: Week07_ANOVA_OneWay/data/

Naming Convention

Dataset filenames follow the pattern: species_trait.csv

Examples: - dairy_lactation.csv - Dairy cattle lactation data - beef_carcass_marbling.csv - Beef carcass quality - swine_growth_breed_diet.csv - Swine growth factorial - broiler_growth.csv - Poultry growth curve

File Format

Format: CSV (comma-separated values) with headers
Missing data: Coded as NA
Encoding: UTF-8
Line endings: Unix-style (LF)

Variable Naming

Variables use descriptive names with underscores: - IDs: cow_id, pig_id, lamb_id - Traits: milk_yield_kg, adg_kg_day, body_weight_kg - Factors: breed, diet, sex, strain

Accessing Datasets

From the book’s root directory:

data <- read.csv("Week04_SimpleRegression/data/broiler_growth.csv")

Or with relative paths from within a week’s folder:

data <- read.csv("data/broiler_growth.csv")

Week 1: Course Overview & Computational Foundations

Statistical Focus: Matrix notation, computing sample mean as a linear model.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`dairy_milk_practice.csv`	Dairy	30	cow_id, milk_yield	Milk yield (kg/day)	Introduction to vectors and matrices

Variable Details:

dairy_milk_practice.csv: - cow_id: Unique cow identifier (1-30) - milk_yield: Daily milk production in kilograms

Pedagogical Design

This is the first dataset students encounter. It demonstrates: - How to represent data as vectors (y) - Building the simplest design matrix (X = column of ones) - Computing the sample mean using matrix operations: $\bar{y} = (\mathbf{X}'\mathbf{X})^{-1}\mathbf{X}'\mathbf{y}$ - Introduction to Holstein dairy production (realistic milk yields 20-35 kg/day)

Week 2: Linear Algebra Essentials

Statistical Focus: Matrix rank, determinants, inverses, generalized inverses, solving systems.

Week 2: Pure Theory

Week 2 focuses exclusively on linear algebra essentials. All examples are inline in the lecture notes with small matrices (typically 2×2 or 3×3) for hand calculations. No CSV files are provided as students work through theoretical concepts.

Topics covered: Rank, linear independence, regular vs. generalized inverse, eigenvalues, QR decomposition.

Week 3: Building the Design Matrix Framework

Statistical Focus: Construct design matrices from raw data, understand coding schemes (cell means, effects model).

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`pig_litter_breeds.csv`	Swine	6	sow_id, breed, litter_size	Litter size by breed	Hand-calculable, cell means model
`broiler_bodyweight_sex.csv`	Poultry	20	bird_id, sex, body_weight_kg	Body weight by sex	Realistic categorical predictor

Variable Details:

pig_litter_breeds.csv: - sow_id: Individual sow identifier - breed: Yorkshire, Landrace, or Duroc (3 levels) - litter_size: Number of piglets born alive

broiler_bodyweight_sex.csv: - bird_id: Individual broiler identifier - sex: Male or Female (2 levels) - body_weight_kg: Body weight in kilograms at processing

Biological Context

Litter size is a key reproductive trait in swine breeding. Genetic differences exist among breeds: - Yorkshire: Known for large litters (average 11-12 pigs) - Landrace: Similar to Yorkshire (10-11 pigs) - Duroc: Slightly smaller litters (9-10 pigs) but higher growth rate

Body weight in broilers shows strong sexual dimorphism, with males typically 15-20% heavier than females at the same age.

Week 4: Simple Linear Regression

Statistical Focus: Fit bivariate models, interpret slope and intercept, compute residuals, make predictions.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`broiler_growth.csv`	Poultry	4	age_days, weight_kg	Weight vs age	Hand-calculable, perfect for learning X’X, X’y
`sheep_weight.csv`	Sheep	5	lamb_id, birth_weight_kg, weaning_weight_kg	Weaning vs birth weight	Small hand-calculable regression
`pig_feed_efficiency.csv`	Swine	20	pig_id, feed_intake_kg, adg_kg_day	ADG vs feed intake	Growth response to nutrition
`dairy_lactation.csv`	Dairy	30	cow_id, days_in_milk, milk_yield_kg	Milk yield vs DIM	Realistic lactation curve (linear approx)

Variable Details:

broiler_growth.csv: - age_days: Age in days (7, 14, 21, 28) - weight_kg: Body weight in kilograms

sheep_weight.csv: - lamb_id: Individual lamb identifier - birth_weight_kg: Weight at birth in kilograms - weaning_weight_kg: Weight at weaning (~60 days) in kilograms

pig_feed_efficiency.csv: - pig_id: Individual pig identifier - feed_intake_kg: Daily feed consumption in kilograms - adg_kg_day: Average daily gain in kg per day

dairy_lactation.csv: - cow_id: Unique cow identifier - days_in_milk: Days since calving (lactation stage) - milk_yield_kg: Daily milk production in kilograms

Pedagogical Design

Week 4 provides datasets of increasing complexity: 1. broiler_growth.csv (n=4) - Perfect for hand calculations to verify lm() output 2. sheep_weight.csv (n=5) - Still hand-calculable but slightly larger 3. pig_feed_efficiency.csv (n=20) - Transition to realistic sample sizes 4. dairy_lactation.csv (n=30) - Authentic livestock production scenario

This progression helps students build confidence from manual computation to software-based analysis.

File Location

All Week 4 datasets are located in: Week04_SimpleRegression/data/

Week 5: Least Squares Theory

Statistical Focus: Derive LS estimates, prove Gauss-Markov theorem, partition sums of squares, compute confidence intervals.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`lamb_data.csv`	Sheep	40	lamb_id, birth_weight_kg, dam_age, weaning_weight_kg	Weaning weight prediction	Full ANOVA table, confidence intervals

Variable Details:

lamb_data.csv: - lamb_id: Individual lamb identifier - birth_weight_kg: Weight at birth in kilograms - dam_age: Age of dam (mother) in years - weaning_weight_kg: Weight at weaning in kilograms

Biological Context

Weaning weight in sheep is influenced by: - Birth weight: Strong positive correlation (heavier birth → heavier weaning) - Dam age: Mature ewes (3-6 years) typically raise heavier lambs than young or old ewes - Genetics: High heritability (~0.15-0.25)

Week 6: Multiple Regression

Statistical Focus: Extend to multiple predictors, interpret partial regression coefficients, assess collinearity, compute R².

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`beef_carcass_marbling.csv`	Beef	50	steer_id, live_weight_kg, ribeye_area_cm2, backfat_mm, marbling_score	Marbling quality	Three predictors, check VIF

Variable Details:

beef_carcass_marbling.csv: - steer_id: Individual steer identifier - live_weight_kg: Live weight before slaughter in kilograms - ribeye_area_cm2: Ribeye muscle area in square centimeters - backfat_mm: Backfat thickness in millimeters - marbling_score: Intramuscular fat score (1-10 scale)

Biological Context

Marbling (intramuscular fat) is a premium carcass quality trait: - Positively correlated with palatability and meat tenderness - Influenced by genetics (heritability ~0.40), nutrition, and management - Predictors: Live weight (growth), ribeye area (muscularity), backfat (overall fatness) - Important for understanding multicollinearity as these predictors are correlated

Week 7: Analysis of Variance (One-Way)

Statistical Focus: Express ANOVA as linear model, partition variation, conduct F-tests, compare groups.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`dairy_milk_breeds.csv`	Dairy	12	breed, milk_yield	Milk yield by breed	Hand-calculable, 4 breeds, balanced (3 per breed)
`broiler_feed_efficiency.csv`	Poultry	50	pen_id, diet, fcr	FCR by diet	Realistic, 5 diets, balanced (10 per diet)
`litter_size_genetics.csv`	Swine	32	sow_id, line, litter_size	Litter size by line	Moderate size, 4 lines, balanced
`swine_growth_unbalanced.csv`	Swine	29	pig_id, diet, adg	ADG by diet	First unbalanced dataset, introduces Type I/II/III SS

Variable Details:

dairy_milk_breeds.csv: - breed: Holstein, Jersey, Brown Swiss, or Ayrshire (4 levels) - milk_yield: Daily milk production in kilograms

broiler_feed_efficiency.csv: - pen_id: Pen identifier - diet: Dietary program (1-5) - fcr: Feed conversion ratio (kg feed / kg gain)

litter_size_genetics.csv: - sow_id: Individual sow identifier - line: Genetic line (A, B, C, D) - litter_size: Number of piglets born alive

swine_growth_unbalanced.csv: - pig_id: Individual pig identifier - diet: Dietary treatment (5 levels) - adg: Average daily gain in kg per day

Special Features

swine_growth_unbalanced.csv intentionally has unequal sample sizes across diets (ranging from 4-8 pigs per diet). This introduces students to: - Challenges of unbalanced designs - Difference between Type I (sequential) and Type III (partial) sums of squares - Loss of orthogonality in ANOVA

Week 8: Contrasts and Estimable Functions

Statistical Focus: Define contrasts, test specific hypotheses, construct orthogonal contrasts, determine estimability.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`beef_breeds_small.csv`	Beef	12	breed, carcass_weight_kg	Carcass weight by breed	Hand-calculable contrasts, 4 breeds, balanced
`layer_strains_egg.csv`	Poultry	50	hen_id, strain, eggs_month	Egg production by strain	Test orthogonal contrasts, 5 strains

Variable Details:

beef_breeds_small.csv: - breed: Angus, Hereford, Charolais, or Simmental (4 levels) - carcass_weight_kg: Hot carcass weight in kilograms

layer_strains_egg.csv: - hen_id: Individual hen identifier - strain: Layer strain/line (5 levels) - eggs_month: Number of eggs produced per month

Example Contrasts

beef_breeds_small.csv enables contrasts such as: - British breeds vs. Continental breeds: (Angus + Hereford)/2 - (Charolais + Simmental)/2 - Angus vs. Hereford: Angus - Hereford - Charolais vs. Simmental: Charolais - Simmental

Students verify orthogonality and partition SS(Breeds) using these contrasts.

Week 9: Two-Way ANOVA and Factorial Models

Statistical Focus: Model main effects and interactions, interpret interaction plots, understand Type I/II/III SS.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`swine_growth_breed_diet.csv`	Swine	8	breed, diet, adg	ADG (breed × diet)	Hand-calculable 2×2 factorial
`lamb_growth_breed_diet.csv`	Sheep	12	lamb_id, breed, diet, weight_gain_kg	Weight gain	Two-way with interaction
`dairy_milk_fat_breed_diet.csv`	Dairy	54	cow_id, breed, diet, fat_pct, protein_pct, days_in_milk	Milk composition	Realistic factorial, multiple responses

Variable Details:

swine_growth_breed_diet.csv: - breed: Yorkshire or Duroc (2 levels) - diet: Diet 1 or Diet 2 (2 levels) - adg: Average daily gain in kg per day

lamb_growth_breed_diet.csv: - lamb_id: Individual lamb identifier - breed: Breed type (3 levels) - diet: Dietary treatment (2 levels) - weight_gain_kg: Weight gain from birth to weaning in kilograms

dairy_milk_fat_breed_diet.csv: - cow_id: Unique cow identifier - breed: Dairy breed (3 levels) - diet: Dietary treatment (2 levels) - fat_pct: Milk fat percentage - protein_pct: Milk protein percentage - days_in_milk: Days since calving

Interaction Interpretation

If breed × diet interaction is significant in swine_growth_breed_diet.csv, it indicates: - The effect of diet on ADG depends on breed - Or equivalently, the breed difference varies by diet - Cannot make general statements about diet or breed effects - Must report simple effects (diet effect within each breed)

Week 10: Analysis of Covariance (ANCOVA)

Statistical Focus: Combine categorical and continuous predictors, adjust treatment means for covariates, test homogeneity of slopes.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`swine_litter_parity.csv`	Swine	9	sow_id, breed, parity, litter_size	Litter size adjusted for parity	Hand-calculable ANCOVA
`layer_egg_bodyweight.csv`	Poultry	30	hen_id, strain, body_weight_kg, eggs_month	Eggs adjusted for body weight	Covariate adjustment
`beef_feedlot_adg.csv`	Beef	40	steer_id, ration, initial_weight_kg, adg	ADG adjusted for initial weight	Remove confounding
`dairy_milk_herds.csv`	Dairy	40	cow_id, herd, days_in_milk, milk_yield_kg	Milk adjusted for DIM	Multi-herd comparison

Variable Details:

swine_litter_parity.csv: - sow_id: Individual sow identifier - breed: Breed (3 levels) - parity: Parity number (reproductive cycle: 1, 2, or 3+) - litter_size: Number of piglets born alive

layer_egg_bodyweight.csv: - hen_id: Individual hen identifier - strain: Layer strain (3 levels) - body_weight_kg: Hen body weight in kilograms - eggs_month: Eggs produced per month

beef_feedlot_adg.csv: - steer_id: Individual steer identifier - ration: Feedlot ration (4 levels) - initial_weight_kg: Weight at feedlot entry in kilograms - adg: Average daily gain in kg per day

dairy_milk_herds.csv: - cow_id: Unique cow identifier - herd: Herd/farm identifier (4 levels) - days_in_milk: Days since calving (covariate) - milk_yield_kg: Daily milk production in kilograms

Covariate Purpose

ANCOVA removes confounding: - Parity affects litter size (mature sows have larger litters) - Body weight affects egg production (heavier hens may lay more eggs) - Initial weight affects ADG (lighter calves may have compensatory gain) - Days in milk strongly affects yield (lactation curve)

Adjusting for these covariates provides fairer comparisons of treatments.

Week 11: Model Diagnostics

Statistical Focus: Check assumptions, identify outliers and influential observations, use diagnostic plots, apply transformations.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`beef_marbling.csv`	Beef	9	obs, live_weight_kg, marbling_score	Marbling vs weight	Hand-calculable, includes outlier
`beef_feedlot_gain.csv`	Beef	45	steer_id, ration, adg	ADG by ration	Residual analysis practice
`broiler_bodyweight_growth.csv`	Poultry	50	bird_id, age_days, weight_kg	Growth curve	Diagnostic patterns
`dairy_lactation_variance.csv`	Dairy	40	cow_id, days_in_milk, milk_yield_kg	Milk vs DIM	Heteroscedasticity example
`dairy_scc_mastitis.csv`	Dairy	100	cow_id, days_in_milk, scc	Somatic cell count	Outliers, influential points
`swine_birth_weight_outliers.csv`	Swine	40	pig_id, litter_id, birth_weight_kg	Birth weight	Intentional outliers
`lamb_weaning_weight.csv`	Sheep	60	lamb_id, sex, birth_weight_kg, weaning_weight_kg	Weaning weight	Comprehensive diagnostics
`poultry_fcr.csv`	Poultry	50	bird_id, fcr, feed_intake_kg, weight_gain_kg	Feed conversion	FCR diagnostics

Variable Details:

beef_marbling.csv: - obs: Observation number - live_weight_kg: Live weight in kilograms - marbling_score: Marbling score (1-10) - Special: One observation is an extreme outlier (very heavy steer, low marbling)

dairy_lactation_variance.csv: - cow_id: Unique cow identifier - days_in_milk: Days since calving - milk_yield_kg: Daily milk yield in kilograms - Special: Variance increases with DIM (heteroscedasticity)

swine_birth_weight_outliers.csv: - pig_id: Individual pig identifier - litter_id: Litter identifier - birth_weight_kg: Birth weight in kilograms - Special: Contains 2-3 intentionally extreme values for outlier detection practice

dairy_scc_mastitis.csv: - cow_id: Unique cow identifier - days_in_milk: Days since calving - scc: Somatic cell count (cells/mL, mastitis indicator) - Special: Several high-leverage, high-influence observations

Week 11 Design Philosophy

Week 11 datasets are specifically designed with diagnostic challenges: - Outliers: Unusual observations far from the regression line - Heteroscedasticity: Non-constant variance - Influential points: High leverage combined with large residuals - Nonlinearity: Patterns in residual plots

These features are intentional to teach students how to recognize and handle violations of model assumptions.

Week 12: Unequal Subclass Numbers & Non-Full Rank Models

Statistical Focus: Handle unbalanced data, work with non-full rank design matrices, use generalized inverses, determine estimability.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`sheep_fleece_unbalanced.csv`	Sheep	6	breed, fleece_weight	Fleece weight	Hand-calculable, unequal n (3, 2, 1)
`dairy_herds_unbalanced.csv`	Dairy	45	cow_id, herd, breed, milk_yield_kg	Milk yield	Multi-herd, unbalanced
`lamb_breed_environment.csv`	Sheep	72	lamb_id, breed, environment, weaning_weight_kg	Weaning weight	Nested design
`beef_multifarm_unbalanced.csv`	Beef	177	steer_id, farm, breed, adg, initial_weight_kg	ADG multi-farm	Largest dataset, missing breed × farm cells

Variable Details:

sheep_fleece_unbalanced.csv: - breed: Sheep breed (3 levels) - fleece_weight: Annual fleece weight in kilograms - Structure: Breed 1 (n=3), Breed 2 (n=2), Breed 3 (n=1)

dairy_herds_unbalanced.csv: - cow_id: Unique cow identifier - herd: Herd identifier (5 herds) - breed: Dairy breed (3 breeds) - milk_yield_kg: Daily milk production - Structure: Not all herds have all breeds

lamb_breed_environment.csv: - lamb_id: Individual lamb identifier - breed: Breed (3 levels) - environment: Production environment (3 levels) - weaning_weight_kg: Weight at weaning - Structure: Breeds nested within environments

beef_multifarm_unbalanced.csv: - steer_id: Individual steer identifier - farm: Farm/feedlot identifier (5 farms) - breed: Beef breed (4 breeds) - adg: Average daily gain - initial_weight_kg: Weight at feedlot entry - Structure: Missing breed × farm combinations, highly unbalanced

Rank Deficiency

Week 12 datasets demonstrate: - Unequal sample sizes: Loss of orthogonality - Missing cells: Design matrix X is not full rank - Normal equations: X’X is singular, requires generalized inverse - Estimability: Individual parameters may not be estimable, but contrasts often are - Type III SS: Essential for unbalanced data

Week 13: Special Topics I

Statistical Focus: Handle constraints, check estimability, preview genetic evaluation concepts (unequal progeny).

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`swine_litter.csv`	Swine	31	sow_id, dam_id, litter_size	Litter size	Unequal progeny per dam
`layer_egg_production.csv`	Poultry	45	hen_id, sire_id, eggs_month	Egg production	Sire comparison, unequal daughters
`beef_feedlot.csv`	Beef	162	steer_id, lot, breed, diet, adg, backfat_mm	Multi-trait feedlot	Complex multi-factor design
`dairy_sire.csv`	Dairy	95	daughter_id, sire_id, sire_name, milk_yield	Milk yield	Genetic evaluation, unequal daughters (3-25 per sire)
`broiler_bodyweight.csv`	Poultry	89	bird_id, strain, sex, body_weight_kg	Body weight	Multi-strain comparison

Variable Details:

dairy_sire.csv: - daughter_id: Individual daughter (cow) identifier - sire_id: Sire identifier (10 sires) - sire_name: Sire name/code - milk_yield: Daughter’s milk production in kilograms - Structure: Highly unbalanced (3 to 25 daughters per sire)

beef_feedlot.csv: - steer_id: Individual steer identifier - lot: Feedlot pen/lot identifier - breed: Beef breed (3 breeds) - diet: Dietary treatment (4 diets) - adg: Average daily gain - backfat_mm: Backfat thickness at slaughter

Genetic Evaluation Preview

dairy_sire.csv introduces concepts critical for animal breeding: - Unequal information: Some sires have many daughters (high reliability), others have few (low reliability) - Shrinkage: Sires with few daughters are “shrunk” toward the population mean - Best Linear Unbiased Prediction (BLUP): Preview of Week 14 mixed models - Estimability: Sire differences are estimable even though individual sire effects are not

Week 14: Special Topics II

Statistical Focus: Polynomial regression, weighted least squares, no-intercept models, mixed model preview.

Dataset Name	Species	n	Variables	Trait(s)	Purpose/Features
`broiler_growth.csv`	Poultry	6	age_days, weight_kg	Growth curve	Reused from Week 4, polynomial regression
`lactation_curve.csv`	Dairy	60	cow_id, days_in_milk, milk_yield_kg	Lactation curve	2nd/3rd degree polynomials
`pen_avg_adg.csv`	Swine	5	pen, adg, pen_size	Pen average ADG	Hand-calculable WLS example
`feed_efficiency_pens.csv`	Swine/Poultry	30	pen_id, diet, fcr, pen_size	FCR by pen	WLS, heterogeneous variance
`milk_feed.csv`	Dairy	5	feed_intake_kg, milk_yield_kg	Milk vs feed	Through-origin regression
`broiler_gain_feed.csv`	Poultry	40	bird_id, feed_intake_kg, weight_gain_kg	Feed efficiency	Gain vs feed
`simple_sire_model.csv`	Dairy	20	daughter_id, sire_id, milk_yield	Milk by sire	Mixed model preview, 3-25 daughters
`dairy_sire_eval.csv`	Dairy	100	daughter_id, sire_id, herd, milk_yield_kg	Sire evaluation	Comprehensive mixed model preview

Variable Details:

lactation_curve.csv: - cow_id: Unique cow identifier - days_in_milk: Days since calving (0-305) - milk_yield_kg: Daily milk yield - Purpose: Fit polynomial models (quadratic, cubic) to capture lactation curve shape

pen_avg_adg.csv: - pen: Pen identifier - adg: Pen average daily gain - pen_size: Number of pigs in pen - Purpose: WLS with weights = pen_size (variance ∝ 1/n)

milk_feed.csv: - feed_intake_kg: Daily feed consumption - milk_yield_kg: Daily milk production - Purpose: No-intercept model (biological constraint: zero feed → zero milk)

dairy_sire_eval.csv: - daughter_id: Individual daughter identifier - sire_id: Sire identifier (10 sires) - herd: Herd identifier (contemporary group) - milk_yield_kg: Daughter’s milk yield - Purpose: Full mixed model with fixed (herd) and random (sire) effects

Dataset Reuse

broiler_growth.csv first appeared in Week 4 for simple linear regression. In Week 14, the same data demonstrates: - Linear model (Week 4): weight = β₀ + β₁(age) - Quadratic model (Week 14): weight = β₀ + β₁(age) + β₂(age²) - Cubic model (Week 14): weight = β₀ + β₁(age) + β₂(age²) + β₃(age³)

This shows how the same data can be analyzed with different models depending on the research question.

Week 15: Capstone Project

Statistical Focus: Integrate all concepts from Weeks 1-14 in a comprehensive multi-breed, multi-farm beef cattle analysis.

Week 15: Dataset Under Development

The Week 15 capstone dataset is planned but not yet created. The planned structure:

Dataset: beef_capstone.csv - Species: Beef cattle (steers) - n: ~120 observations - Breeds: Angus, Hereford, Charolais (3 levels) - Farms: 5 feedlots (unbalanced design) - Traits: - adg: Average daily gain (kg/day) - initial_weight_kg: Weight at feedlot entry - days_on_feed: Duration of feedlot period - Design Features: - Missing breed × farm cells (not all breeds on all farms) - Unbalanced sample sizes (4-15 steers per cell) - Intentional outliers for diagnostic practice - Covariate (initial weight) for ANCOVA - Rank deficiency requiring generalized inverse

Analysis Tasks: 1. Exploratory data analysis and design matrix construction 2. Multiple regression with covariates 3. Two-way ANOVA (breed × farm) with interaction testing 4. Model diagnostics (outliers, leverage, influence) 5. Handle rank deficiency and estimability 6. Build custom least squares solver from scratch 7. Compare manual calculations with lm() output

Students integrate all course concepts in this comprehensive real-world scenario.

Species Index

Datasets grouped by livestock species for easy lookup.

Dairy Cattle (13 datasets)

Dataset	Week	n	Traits
dairy_milk_practice.csv	1	30	Milk yield
dairy_lactation.csv	4	30	Milk vs DIM
dairy_milk_breeds.csv	7	12	Milk by breed
dairy_milk_fat_breed_diet.csv	9	54	Milk composition, factorial
dairy_milk_herds.csv	10	40	Milk by herd, adjust for DIM
dairy_lactation_variance.csv	11	40	Heteroscedasticity example
dairy_scc_mastitis.csv	11	100	SCC, outliers/influence
dairy_herds_unbalanced.csv	12	45	Multi-herd unbalanced
dairy_sire.csv	13	95	Sire evaluation, unequal progeny
lactation_curve.csv	14	60	Polynomial lactation curve
milk_feed.csv	14	5	Through-origin regression
simple_sire_model.csv	14	20	Mixed model preview
dairy_sire_eval.csv	14	100	Comprehensive sire evaluation

Beef Cattle (9 datasets)

Dataset	Week	n	Traits
beef_carcass_marbling.csv	6	50	Marbling, multiple predictors
beef_breeds_small.csv	8	12	Carcass weight, contrasts
beef_feedlot_adg.csv	10	40	ADG, ANCOVA
beef_marbling.csv	11	9	Marbling with outlier
beef_feedlot_gain.csv	11	45	ADG, diagnostics
beef_multifarm_unbalanced.csv	12	177	Largest dataset, missing cells
beef_feedlot.csv	13	162	Multi-factor feedlot
beef_capstone.csv (planned)	15	~120	Comprehensive capstone

Swine (8 datasets)

Dataset	Week	n	Traits
pig_litter_breeds.csv	3	6	Litter size by breed
pig_feed_efficiency.csv	4	20	ADG vs feed intake
litter_size_genetics.csv	7	32	Litter size by line
swine_growth_unbalanced.csv	7	29	ADG unbalanced
swine_growth_breed_diet.csv	9	8	Factorial 2×2
swine_litter_parity.csv	10	9	Litter size ANCOVA
swine_birth_weight_outliers.csv	11	40	Intentional outliers
swine_litter.csv	13	31	Unequal progeny
pen_avg_adg.csv	14	5	WLS example
feed_efficiency_pens.csv	14	30	WLS heterogeneous variance

Poultry (12 datasets)

Dataset	Week	n	Traits
broiler_bodyweight_sex.csv	3	20	Weight by sex
broiler_growth.csv	4, 14	4-6	Growth curve (reused)
broiler_feed_efficiency.csv	7	50	FCR by ration
layer_strains_egg.csv	8	50	Egg production
layer_egg_bodyweight.csv	10	30	Eggs ANCOVA
broiler_bodyweight_growth.csv	11	50	Growth diagnostics
poultry_fcr.csv	11	50	FCR diagnostics
layer_egg_production.csv	13	45	Sire comparison
broiler_bodyweight.csv	13	89	Multi-strain
broiler_gain_feed.csv	14	40	Feed efficiency

Sheep (6 datasets)

Dataset	Week	n	Traits
sheep_weight.csv	4	5	Birth to weaning weight
lamb_data.csv	5	40	Weaning weight prediction
lamb_growth_breed_diet.csv	9	12	Factorial growth
lamb_weaning_weight.csv	11	60	Comprehensive diagnostics
sheep_fleece_unbalanced.csv	12	6	Fleece weight unbalanced
lamb_breed_environment.csv	12	72	Nested design

Dataset Size Guide

Datasets organized by sample size to help find appropriate examples for different learning goals.

Hand-Calculable (4-12 observations)

Perfect for manual matrix calculations, verifying software output, and understanding mechanics.

Dataset	Week	n	Species	Purpose
broiler_growth.csv	4	4	Poultry	Simple regression by hand
sheep_weight.csv	4	5	Sheep	Regression practice
milk_feed.csv	14	5	Dairy	Through-origin
pen_avg_adg.csv	14	5	Swine	WLS by hand
pig_litter_breeds.csv	3	6	Swine	Design matrix
sheep_fleece_unbalanced.csv	12	6	Sheep	Unbalanced by hand
swine_growth_breed_diet.csv	9	8	Swine	2×2 factorial
beef_marbling.csv	11	9	Beef	Outlier detection
swine_litter_parity.csv	10	9	Swine	ANCOVA by hand
beef_breeds_small.csv	8	12	Beef	Contrasts
dairy_milk_breeds.csv	7	12	Dairy	One-way ANOVA
lamb_growth_breed_diet.csv	9	12	Sheep	Factorial

Moderate (20-60 observations)

Realistic sample sizes for most livestock experiments, transition from hand calculation to software.

Dataset	Week	n	Species
pig_feed_efficiency.csv	4	20	Swine
broiler_bodyweight_sex.csv	3	20	Poultry
simple_sire_model.csv	14	20	Dairy
swine_growth_unbalanced.csv	7	29	Swine
dairy_milk_practice.csv	1	30	Dairy
dairy_lactation.csv	4	30	Dairy
layer_egg_bodyweight.csv	10	30	Poultry
feed_efficiency_pens.csv	14	30	Swine
swine_litter.csv	13	31	Swine
litter_size_genetics.csv	7	32	Swine
lamb_data.csv	5	40	Sheep
beef_feedlot_adg.csv	10	40	Beef
dairy_milk_herds.csv	10	40	Dairy
dairy_lactation_variance.csv	11	40	Dairy
swine_birth_weight_outliers.csv	11	40	Swine
broiler_gain_feed.csv	14	40	Poultry
dairy_herds_unbalanced.csv	12	45	Dairy
beef_feedlot_gain.csv	11	45	Beef
layer_egg_production.csv	13	45	Poultry
beef_carcass_marbling.csv	6	50	Beef
broiler_feed_efficiency.csv	7	50	Poultry
layer_strains_egg.csv	8	50	Poultry
broiler_bodyweight_growth.csv	11	50	Poultry
poultry_fcr.csv	11	50	Poultry
dairy_milk_fat_breed_diet.csv	9	54	Dairy
lactation_curve.csv	14	60	Dairy
lamb_weaning_weight.csv	11	60	Sheep

Large (60+ observations)

Complex datasets for advanced analyses, diagnostics, and realistic production scenarios.

Dataset	Week	n	Species	Features
lamb_breed_environment.csv	12	72	Sheep	Nested design
broiler_bodyweight.csv	13	89	Poultry	Multi-strain
dairy_sire.csv	13	95	Dairy	Genetic evaluation
dairy_scc_mastitis.csv	11	100	Dairy	Outliers/influence
dairy_sire_eval.csv	14	100	Dairy	Mixed model
beef_feedlot.csv	13	162	Beef	Multi-factor complex
beef_multifarm_unbalanced.csv	12	177	Beef	Largest, missing cells

Quick Reference: Alphabetical Listing

All 48 datasets in alphabetical order with key information.

Dataset	Week	Species	n	Primary Trait(s)
beef_breeds_small.csv	8	Beef	12	Carcass weight
beef_capstone.csv (planned)	15	Beef	~120	ADG multi-factor
beef_carcass_marbling.csv	6	Beef	50	Marbling quality
beef_feedlot.csv	13	Beef	162	Multi-trait feedlot
beef_feedlot_adg.csv	10	Beef	40	ADG ANCOVA
beef_feedlot_gain.csv	11	Beef	45	ADG diagnostics
beef_marbling.csv	11	Beef	9	Marbling with outlier
beef_multifarm_unbalanced.csv	12	Beef	177	ADG unbalanced
broiler_bodyweight.csv	13	Poultry	89	Body weight multi-strain
broiler_bodyweight_growth.csv	11	Poultry	50	Growth diagnostics
broiler_bodyweight_sex.csv	3	Poultry	20	Weight by sex
broiler_feed_efficiency.csv	7	Poultry	50	FCR by ration
broiler_gain_feed.csv	14	Poultry	40	Feed efficiency
broiler_growth.csv	4, 14	Poultry	4-6	Growth curve
dairy_herds_unbalanced.csv	12	Dairy	45	Milk unbalanced
dairy_lactation.csv	4	Dairy	30	Lactation curve
dairy_lactation_variance.csv	11	Dairy	40	Heteroscedasticity
dairy_milk_breeds.csv	7	Dairy	12	Milk by breed
dairy_milk_fat_breed_diet.csv	9	Dairy	54	Milk composition
dairy_milk_herds.csv	10	Dairy	40	Milk by herd
dairy_milk_practice.csv	1	Dairy	30	Milk yield
dairy_scc_mastitis.csv	11	Dairy	100	SCC outliers
dairy_sire.csv	13	Dairy	95	Sire evaluation
dairy_sire_eval.csv	14	Dairy	100	Mixed model sire
feed_efficiency_pens.csv	14	Swine	30	FCR WLS
lactation_curve.csv	14	Dairy	60	Polynomial lactation
lamb_breed_environment.csv	12	Sheep	72	Nested design
lamb_data.csv	5	Sheep	40	Weaning weight
lamb_growth_breed_diet.csv	9	Sheep	12	Factorial growth
lamb_weaning_weight.csv	11	Sheep	60	Diagnostics
layer_egg_bodyweight.csv	10	Poultry	30	Eggs ANCOVA
layer_egg_production.csv	13	Poultry	45	Sire comparison
layer_strains_egg.csv	8	Poultry	50	Egg production
litter_size_genetics.csv	7	Swine	32	Litter size
milk_feed.csv	14	Dairy	5	Through-origin
pen_avg_adg.csv	14	Swine	5	WLS pens
pig_feed_efficiency.csv	4	Swine	20	ADG vs feed
pig_litter_breeds.csv	3	Swine	6	Litter size
poultry_fcr.csv	11	Poultry	50	FCR diagnostics
sheep_fleece_unbalanced.csv	12	Sheep	6	Fleece weight
sheep_weight.csv	4	Sheep	5	Birth to weaning
simple_sire_model.csv	14	Dairy	20	Mixed model preview
swine_birth_weight_outliers.csv	11	Swine	40	Outliers
swine_growth_breed_diet.csv	9	Swine	8	Factorial 2×2
swine_growth_unbalanced.csv	7	Swine	29	ADG unbalanced
swine_litter.csv	13	Swine	31	Unequal progeny
swine_litter_parity.csv	10	Swine	9	Litter ANCOVA

Notes and Future Plans

Missing Data Conventions

Missing values are coded as NA in all CSV files
Students should check for missing data before analysis
Missing data is rare in these datasets (most are complete cases)
When present, missingness is documented in variable descriptions

Data Integrity

All datasets have been verified for: - Correct variable names matching descriptions - Realistic trait values for each species - Appropriate sample sizes for pedagogical purposes - Consistent units (kg, days, etc.)

Accessing Data Files

All datasets are available in week-specific subdirectories: - Path pattern: WeekXX_TopicName/data/dataset_name.csv - Files are plain-text CSV with Unix line endings - First row contains variable names (headers) - No row names included

Future Dataset Development

Week 15 Capstone Dataset: - Currently under development - Will provide comprehensive integration of all course concepts - Designed to challenge students with realistic complexity - Includes all major features: unbalanced design, covariates, outliers, rank deficiency

Potential Additions: - Additional aquaculture datasets (tilapia, salmon) - Horse breeding examples - Multi-trait genetic evaluation datasets - Longitudinal/repeated measures examples

Citation and Use

These datasets were created specifically for educational purposes in this linear models course. While inspired by real livestock production systems, all data values are simulated to ensure appropriate pedagogical features. Students may use these datasets for practice and learning but should cite this course if using them in reports or presentations.

Feedback

If you encounter issues with any dataset (incorrect values, unclear variables, missing files), please report to the course instructor. Continuous improvement of these educational resources is a priority.

This completes Appendix D. For mathematical notation, see Appendix A. For R function reference, see Appendix C.

# Appendix D: Datasets Description {#sec-appendix-d .unnumbered} ## Introduction This appendix provides a comprehensive reference guide to all datasets used throughout the 15-week linear models course. Each dataset is carefully designed to illustrate specific statistical concepts while maintaining biological realism relevant to animal breeding and genetics. ### Purpose and Scope This reference is designed to: - **Document all datasets**: Complete descriptions of 48 datasets spanning 5 livestock species - **Provide context**: Explain the pedagogical purpose and biological relevance of each dataset - **Enable quick lookup**: Multiple access paths (by week, species, size, or alphabetically) - **Support learning**: Variable descriptions and special features clearly noted ### How to Use This Appendix - **Week-by-week sections**: Find datasets organized chronologically (Weeks 1-15) - **Species Index** ([@sec-datasets-species]): Browse datasets by livestock type - **Size Guide** ([@sec-datasets-size]): Find hand-calculable vs. realistic datasets - **Quick Reference** ([@sec-datasets-quick-ref]): Alphabetical listing of all datasets ### Relationship to Course Structure Datasets are organized to match the 15-week course progression: - **Weeks 1-3**: Small examples for matrix operations and design matrices - **Weeks 4-6**: Regression datasets (simple and multiple) - **Weeks 7-10**: ANOVA family (one-way, contrasts, two-way, ANCOVA) - **Week 11**: Diagnostic datasets with special features (outliers, influential points) - **Weeks 12-14**: Advanced topics (unbalanced, constraints, special cases) - **Week 15**: Comprehensive capstone integration --- ## Summary Statistics {#sec-datasets-summary} ### Overview | Metric | Value | |--------|-------| | **Total datasets** | 48 | | **Total observations** | 2,011 | | **Size range** | 4-177 observations | | **Species covered** | 5 (Dairy, Beef, Swine, Poultry, Sheep) | | **Weeks with datasets** | 14 (Week 2 is pure theory) | ### Distribution by Week | Week | # Datasets | Total Obs | Statistical Focus | |------|------------|-----------|-------------------| | 1 | 1 | 30 | Matrix notation, sample mean | | 2 | 0 | 0 | Linear algebra theory | | 3 | 2 | 26 | Design matrix construction | | 4 | 4 | 59 | Simple linear regression | | 5 | 1 | 40 | Least squares theory | | 6 | 1 | 50 | Multiple regression | | 7 | 4 | 123 | One-way ANOVA | | 8 | 2 | 62 | Contrasts, estimable functions | | 9 | 3 | 74 | Two-way ANOVA, interactions | | 10 | 4 | 119 | ANCOVA, adjusted means | | 11 | 8 | 425 | Model diagnostics | | 12 | 4 | 300 | Non-full rank, unbalanced | | 13 | 5 | 422 | Special topics, constraints | | 14 | 8 | 300 | Polynomial, WLS, mixed model preview | | 15 | 1 | ~120 (planned) | Capstone integration | ### Distribution by Species | Species | # Datasets | Avg Size | Size Range | Primary Traits | |---------|-----------|----------|------------|----------------| | Dairy Cattle | 13 | 52 | 12-100 | Milk yield, lactation curves, SCC | | Beef Cattle | 9 | 56 | 9-177 | ADG, marbling, carcass traits | | Swine | 8 | 26 | 6-162 | Litter size, ADG, growth | | Poultry | 12 | 41 | 4-89 | Egg production, body weight, FCR | | Sheep | 6 | 32 | 5-72 | Weaning weight, fleece, growth | ### Distribution by Size Category | Category | Size Range | # Datasets | Purpose | Example Weeks | |----------|------------|------------|---------|---------------| | Hand-calculable | 4-12 | 15 | Manual computation practice | 1, 3, 4, 7, 8, 9, 12 | | Moderate | 20-50 | 25 | Realistic applications | 4, 5, 6, 7, 8, 10, 13, 14 | | Large | 60-177 | 8 | Complex structures, diagnostics | 11, 12, 13, 14 | --- ## File Conventions {#sec-datasets-conventions} ### Directory Structure All datasets are stored in week-specific subdirectories: ``` WeekXX_TopicName/data/dataset_name.csv ``` **Example**: - Week 4 datasets: `Week04_SimpleRegression/data/` - Week 7 datasets: `Week07_ANOVA_OneWay/data/` ### Naming Convention Dataset filenames follow the pattern: `species_trait.csv` **Examples**: - `dairy_lactation.csv` - Dairy cattle lactation data - `beef_carcass_marbling.csv` - Beef carcass quality - `swine_growth_breed_diet.csv` - Swine growth factorial - `broiler_growth.csv` - Poultry growth curve ### File Format - **Format**: CSV (comma-separated values) with headers - **Missing data**: Coded as `NA` - **Encoding**: UTF-8 - **Line endings**: Unix-style (LF) ### Variable Naming Variables use descriptive names with underscores: - IDs: `cow_id`, `pig_id`, `lamb_id` - Traits: `milk_yield_kg`, `adg_kg_day`, `body_weight_kg` - Factors: `breed`, `diet`, `sex`, `strain` :::{.callout-tip} ## Accessing Datasets From the book's root directory: ```r data <- read.csv("Week04_SimpleRegression/data/broiler_growth.csv") ``` Or with relative paths from within a week's folder: ```r data <- read.csv("data/broiler_growth.csv") ``` ::: --- ## Week 1: Course Overview & Computational Foundations {#sec-datasets-week1} **Statistical Focus**: Matrix notation, computing sample mean as a linear model. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `dairy_milk_practice.csv` | Dairy | 30 | cow_id, milk_yield | Milk yield (kg/day) | Introduction to vectors and matrices | **Variable Details**: **dairy_milk_practice.csv**: - `cow_id`: Unique cow identifier (1-30) - `milk_yield`: Daily milk production in kilograms :::{.callout-note} ## Pedagogical Design This is the first dataset students encounter. It demonstrates: - How to represent data as vectors (**y**) - Building the simplest design matrix (**X** = column of ones) - Computing the sample mean using matrix operations: $\bar{y} = (\mathbf{X}'\mathbf{X})^{-1}\mathbf{X}'\mathbf{y}$ - Introduction to Holstein dairy production (realistic milk yields 20-35 kg/day) ::: --- ## Week 2: Linear Algebra Essentials {#sec-datasets-week2} **Statistical Focus**: Matrix rank, determinants, inverses, generalized inverses, solving systems. :::{.callout-note} ## Week 2: Pure Theory Week 2 focuses exclusively on linear algebra essentials. All examples are inline in the lecture notes with small matrices (typically 2×2 or 3×3) for hand calculations. No CSV files are provided as students work through theoretical concepts. **Topics covered**: Rank, linear independence, regular vs. generalized inverse, eigenvalues, QR decomposition. ::: --- ## Week 3: Building the Design Matrix Framework {#sec-datasets-week3} **Statistical Focus**: Construct design matrices from raw data, understand coding schemes (cell means, effects model). | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `pig_litter_breeds.csv` | Swine | 6 | sow_id, breed, litter_size | Litter size by breed | Hand-calculable, cell means model | | `broiler_bodyweight_sex.csv` | Poultry | 20 | bird_id, sex, body_weight_kg | Body weight by sex | Realistic categorical predictor | **Variable Details**: **pig_litter_breeds.csv**: - `sow_id`: Individual sow identifier - `breed`: Yorkshire, Landrace, or Duroc (3 levels) - `litter_size`: Number of piglets born alive **broiler_bodyweight_sex.csv**: - `bird_id`: Individual broiler identifier - `sex`: Male or Female (2 levels) - `body_weight_kg`: Body weight in kilograms at processing :::{.callout-note} ## Biological Context **Litter size** is a key reproductive trait in swine breeding. Genetic differences exist among breeds: - Yorkshire: Known for large litters (average 11-12 pigs) - Landrace: Similar to Yorkshire (10-11 pigs) - Duroc: Slightly smaller litters (9-10 pigs) but higher growth rate **Body weight** in broilers shows strong sexual dimorphism, with males typically 15-20% heavier than females at the same age. ::: --- ## Week 4: Simple Linear Regression {#sec-datasets-week4} **Statistical Focus**: Fit bivariate models, interpret slope and intercept, compute residuals, make predictions. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `broiler_growth.csv` | Poultry | 4 | age_days, weight_kg | Weight vs age | Hand-calculable, perfect for learning X'X, X'y | | `sheep_weight.csv` | Sheep | 5 | lamb_id, birth_weight_kg, weaning_weight_kg | Weaning vs birth weight | Small hand-calculable regression | | `pig_feed_efficiency.csv` | Swine | 20 | pig_id, feed_intake_kg, adg_kg_day | ADG vs feed intake | Growth response to nutrition | | `dairy_lactation.csv` | Dairy | 30 | cow_id, days_in_milk, milk_yield_kg | Milk yield vs DIM | Realistic lactation curve (linear approx) | **Variable Details**: **broiler_growth.csv**: - `age_days`: Age in days (7, 14, 21, 28) - `weight_kg`: Body weight in kilograms **sheep_weight.csv**: - `lamb_id`: Individual lamb identifier - `birth_weight_kg`: Weight at birth in kilograms - `weaning_weight_kg`: Weight at weaning (~60 days) in kilograms **pig_feed_efficiency.csv**: - `pig_id`: Individual pig identifier - `feed_intake_kg`: Daily feed consumption in kilograms - `adg_kg_day`: Average daily gain in kg per day **dairy_lactation.csv**: - `cow_id`: Unique cow identifier - `days_in_milk`: Days since calving (lactation stage) - `milk_yield_kg`: Daily milk production in kilograms :::{.callout-note} ## Pedagogical Design Week 4 provides datasets of increasing complexity: 1. **broiler_growth.csv** (n=4) - Perfect for hand calculations to verify `lm()` output 2. **sheep_weight.csv** (n=5) - Still hand-calculable but slightly larger 3. **pig_feed_efficiency.csv** (n=20) - Transition to realistic sample sizes 4. **dairy_lactation.csv** (n=30) - Authentic livestock production scenario This progression helps students build confidence from manual computation to software-based analysis. ::: :::{.callout-tip} ## File Location All Week 4 datasets are located in: `Week04_SimpleRegression/data/` ::: --- ## Week 5: Least Squares Theory {#sec-datasets-week5} **Statistical Focus**: Derive LS estimates, prove Gauss-Markov theorem, partition sums of squares, compute confidence intervals. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `lamb_data.csv` | Sheep | 40 | lamb_id, birth_weight_kg, dam_age, weaning_weight_kg | Weaning weight prediction | Full ANOVA table, confidence intervals | **Variable Details**: **lamb_data.csv**: - `lamb_id`: Individual lamb identifier - `birth_weight_kg`: Weight at birth in kilograms - `dam_age`: Age of dam (mother) in years - `weaning_weight_kg`: Weight at weaning in kilograms :::{.callout-note} ## Biological Context Weaning weight in sheep is influenced by: - **Birth weight**: Strong positive correlation (heavier birth → heavier weaning) - **Dam age**: Mature ewes (3-6 years) typically raise heavier lambs than young or old ewes - **Genetics**: High heritability (~0.15-0.25) ::: --- ## Week 6: Multiple Regression {#sec-datasets-week6} **Statistical Focus**: Extend to multiple predictors, interpret partial regression coefficients, assess collinearity, compute R². | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `beef_carcass_marbling.csv` | Beef | 50 | steer_id, live_weight_kg, ribeye_area_cm2, backfat_mm, marbling_score | Marbling quality | Three predictors, check VIF | **Variable Details**: **beef_carcass_marbling.csv**: - `steer_id`: Individual steer identifier - `live_weight_kg`: Live weight before slaughter in kilograms - `ribeye_area_cm2`: Ribeye muscle area in square centimeters - `backfat_mm`: Backfat thickness in millimeters - `marbling_score`: Intramuscular fat score (1-10 scale) :::{.callout-note} ## Biological Context **Marbling** (intramuscular fat) is a premium carcass quality trait: - Positively correlated with palatability and meat tenderness - Influenced by genetics (heritability ~0.40), nutrition, and management - Predictors: Live weight (growth), ribeye area (muscularity), backfat (overall fatness) - Important for understanding multicollinearity as these predictors are correlated ::: --- ## Week 7: Analysis of Variance (One-Way) {#sec-datasets-week7} **Statistical Focus**: Express ANOVA as linear model, partition variation, conduct F-tests, compare groups. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `dairy_milk_breeds.csv` | Dairy | 12 | breed, milk_yield | Milk yield by breed | Hand-calculable, 4 breeds, balanced (3 per breed) | | `broiler_feed_efficiency.csv` | Poultry | 50 | pen_id, diet, fcr | FCR by diet | Realistic, 5 diets, balanced (10 per diet) | | `litter_size_genetics.csv` | Swine | 32 | sow_id, line, litter_size | Litter size by line | Moderate size, 4 lines, balanced | | `swine_growth_unbalanced.csv` | Swine | 29 | pig_id, diet, adg | ADG by diet | **First unbalanced dataset**, introduces Type I/II/III SS | **Variable Details**: **dairy_milk_breeds.csv**: - `breed`: Holstein, Jersey, Brown Swiss, or Ayrshire (4 levels) - `milk_yield`: Daily milk production in kilograms **broiler_feed_efficiency.csv**: - `pen_id`: Pen identifier - `diet`: Dietary program (1-5) - `fcr`: Feed conversion ratio (kg feed / kg gain) **litter_size_genetics.csv**: - `sow_id`: Individual sow identifier - `line`: Genetic line (A, B, C, D) - `litter_size`: Number of piglets born alive **swine_growth_unbalanced.csv**: - `pig_id`: Individual pig identifier - `diet`: Dietary treatment (5 levels) - `adg`: Average daily gain in kg per day :::{.callout-important} ## Special Features `swine_growth_unbalanced.csv` intentionally has unequal sample sizes across diets (ranging from 4-8 pigs per diet). This introduces students to: - Challenges of unbalanced designs - Difference between Type I (sequential) and Type III (partial) sums of squares - Loss of orthogonality in ANOVA ::: --- ## Week 8: Contrasts and Estimable Functions {#sec-datasets-week8} **Statistical Focus**: Define contrasts, test specific hypotheses, construct orthogonal contrasts, determine estimability. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `beef_breeds_small.csv` | Beef | 12 | breed, carcass_weight_kg | Carcass weight by breed | Hand-calculable contrasts, 4 breeds, balanced | | `layer_strains_egg.csv` | Poultry | 50 | hen_id, strain, eggs_month | Egg production by strain | Test orthogonal contrasts, 5 strains | **Variable Details**: **beef_breeds_small.csv**: - `breed`: Angus, Hereford, Charolais, or Simmental (4 levels) - `carcass_weight_kg`: Hot carcass weight in kilograms **layer_strains_egg.csv**: - `hen_id`: Individual hen identifier - `strain`: Layer strain/line (5 levels) - `eggs_month`: Number of eggs produced per month :::{.callout-note} ## Example Contrasts **beef_breeds_small.csv** enables contrasts such as: - British breeds vs. Continental breeds: (Angus + Hereford)/2 - (Charolais + Simmental)/2 - Angus vs. Hereford: Angus - Hereford - Charolais vs. Simmental: Charolais - Simmental Students verify orthogonality and partition SS(Breeds) using these contrasts. ::: --- ## Week 9: Two-Way ANOVA and Factorial Models {#sec-datasets-week9} **Statistical Focus**: Model main effects and interactions, interpret interaction plots, understand Type I/II/III SS. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `swine_growth_breed_diet.csv` | Swine | 8 | breed, diet, adg | ADG (breed × diet) | Hand-calculable 2×2 factorial | | `lamb_growth_breed_diet.csv` | Sheep | 12 | lamb_id, breed, diet, weight_gain_kg | Weight gain | Two-way with interaction | | `dairy_milk_fat_breed_diet.csv` | Dairy | 54 | cow_id, breed, diet, fat_pct, protein_pct, days_in_milk | Milk composition | Realistic factorial, multiple responses | **Variable Details**: **swine_growth_breed_diet.csv**: - `breed`: Yorkshire or Duroc (2 levels) - `diet`: Diet 1 or Diet 2 (2 levels) - `adg`: Average daily gain in kg per day **lamb_growth_breed_diet.csv**: - `lamb_id`: Individual lamb identifier - `breed`: Breed type (3 levels) - `diet`: Dietary treatment (2 levels) - `weight_gain_kg`: Weight gain from birth to weaning in kilograms **dairy_milk_fat_breed_diet.csv**: - `cow_id`: Unique cow identifier - `breed`: Dairy breed (3 levels) - `diet`: Dietary treatment (2 levels) - `fat_pct`: Milk fat percentage - `protein_pct`: Milk protein percentage - `days_in_milk`: Days since calving :::{.callout-important} ## Interaction Interpretation If breed × diet interaction is significant in `swine_growth_breed_diet.csv`, it indicates: - The effect of diet on ADG **depends on breed** - Or equivalently, the breed difference **varies by diet** - Cannot make general statements about diet or breed effects - Must report simple effects (diet effect within each breed) ::: --- ## Week 10: Analysis of Covariance (ANCOVA) {#sec-datasets-week10} **Statistical Focus**: Combine categorical and continuous predictors, adjust treatment means for covariates, test homogeneity of slopes. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `swine_litter_parity.csv` | Swine | 9 | sow_id, breed, parity, litter_size | Litter size adjusted for parity | Hand-calculable ANCOVA | | `layer_egg_bodyweight.csv` | Poultry | 30 | hen_id, strain, body_weight_kg, eggs_month | Eggs adjusted for body weight | Covariate adjustment | | `beef_feedlot_adg.csv` | Beef | 40 | steer_id, ration, initial_weight_kg, adg | ADG adjusted for initial weight | Remove confounding | | `dairy_milk_herds.csv` | Dairy | 40 | cow_id, herd, days_in_milk, milk_yield_kg | Milk adjusted for DIM | Multi-herd comparison | **Variable Details**: **swine_litter_parity.csv**: - `sow_id`: Individual sow identifier - `breed`: Breed (3 levels) - `parity`: Parity number (reproductive cycle: 1, 2, or 3+) - `litter_size`: Number of piglets born alive **layer_egg_bodyweight.csv**: - `hen_id`: Individual hen identifier - `strain`: Layer strain (3 levels) - `body_weight_kg`: Hen body weight in kilograms - `eggs_month`: Eggs produced per month **beef_feedlot_adg.csv**: - `steer_id`: Individual steer identifier - `ration`: Feedlot ration (4 levels) - `initial_weight_kg`: Weight at feedlot entry in kilograms - `adg`: Average daily gain in kg per day **dairy_milk_herds.csv**: - `cow_id`: Unique cow identifier - `herd`: Herd/farm identifier (4 levels) - `days_in_milk`: Days since calving (covariate) - `milk_yield_kg`: Daily milk production in kilograms :::{.callout-note} ## Covariate Purpose **ANCOVA removes confounding**: - **Parity** affects litter size (mature sows have larger litters) - **Body weight** affects egg production (heavier hens may lay more eggs) - **Initial weight** affects ADG (lighter calves may have compensatory gain) - **Days in milk** strongly affects yield (lactation curve) Adjusting for these covariates provides fairer comparisons of treatments. ::: --- ## Week 11: Model Diagnostics {#sec-datasets-week11} **Statistical Focus**: Check assumptions, identify outliers and influential observations, use diagnostic plots, apply transformations. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `beef_marbling.csv` | Beef | 9 | obs, live_weight_kg, marbling_score | Marbling vs weight | Hand-calculable, **includes outlier** | | `beef_feedlot_gain.csv` | Beef | 45 | steer_id, ration, adg | ADG by ration | Residual analysis practice | | `broiler_bodyweight_growth.csv` | Poultry | 50 | bird_id, age_days, weight_kg | Growth curve | Diagnostic patterns | | `dairy_lactation_variance.csv` | Dairy | 40 | cow_id, days_in_milk, milk_yield_kg | Milk vs DIM | **Heteroscedasticity** example | | `dairy_scc_mastitis.csv` | Dairy | 100 | cow_id, days_in_milk, scc | Somatic cell count | Outliers, influential points | | `swine_birth_weight_outliers.csv` | Swine | 40 | pig_id, litter_id, birth_weight_kg | Birth weight | **Intentional outliers** | | `lamb_weaning_weight.csv` | Sheep | 60 | lamb_id, sex, birth_weight_kg, weaning_weight_kg | Weaning weight | Comprehensive diagnostics | | `poultry_fcr.csv` | Poultry | 50 | bird_id, fcr, feed_intake_kg, weight_gain_kg | Feed conversion | FCR diagnostics | **Variable Details**: **beef_marbling.csv**: - `obs`: Observation number - `live_weight_kg`: Live weight in kilograms - `marbling_score`: Marbling score (1-10) - **Special**: One observation is an extreme outlier (very heavy steer, low marbling) **dairy_lactation_variance.csv**: - `cow_id`: Unique cow identifier - `days_in_milk`: Days since calving - `milk_yield_kg`: Daily milk yield in kilograms - **Special**: Variance increases with DIM (heteroscedasticity) **swine_birth_weight_outliers.csv**: - `pig_id`: Individual pig identifier - `litter_id`: Litter identifier - `birth_weight_kg`: Birth weight in kilograms - **Special**: Contains 2-3 intentionally extreme values for outlier detection practice **dairy_scc_mastitis.csv**: - `cow_id`: Unique cow identifier - `days_in_milk`: Days since calving - `scc`: Somatic cell count (cells/mL, mastitis indicator) - **Special**: Several high-leverage, high-influence observations :::{.callout-important} ## Week 11 Design Philosophy Week 11 datasets are specifically designed with diagnostic challenges: - **Outliers**: Unusual observations far from the regression line - **Heteroscedasticity**: Non-constant variance - **Influential points**: High leverage combined with large residuals - **Nonlinearity**: Patterns in residual plots These features are intentional to teach students how to recognize and handle violations of model assumptions. ::: --- ## Week 12: Unequal Subclass Numbers & Non-Full Rank Models {#sec-datasets-week12} **Statistical Focus**: Handle unbalanced data, work with non-full rank design matrices, use generalized inverses, determine estimability. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `sheep_fleece_unbalanced.csv` | Sheep | 6 | breed, fleece_weight | Fleece weight | Hand-calculable, unequal n (3, 2, 1) | | `dairy_herds_unbalanced.csv` | Dairy | 45 | cow_id, herd, breed, milk_yield_kg | Milk yield | Multi-herd, unbalanced | | `lamb_breed_environment.csv` | Sheep | 72 | lamb_id, breed, environment, weaning_weight_kg | Weaning weight | Nested design | | `beef_multifarm_unbalanced.csv` | Beef | 177 | steer_id, farm, breed, adg, initial_weight_kg | ADG multi-farm | **Largest dataset**, missing breed × farm cells | **Variable Details**: **sheep_fleece_unbalanced.csv**: - `breed`: Sheep breed (3 levels) - `fleece_weight`: Annual fleece weight in kilograms - **Structure**: Breed 1 (n=3), Breed 2 (n=2), Breed 3 (n=1) **dairy_herds_unbalanced.csv**: - `cow_id`: Unique cow identifier - `herd`: Herd identifier (5 herds) - `breed`: Dairy breed (3 breeds) - `milk_yield_kg`: Daily milk production - **Structure**: Not all herds have all breeds **lamb_breed_environment.csv**: - `lamb_id`: Individual lamb identifier - `breed`: Breed (3 levels) - `environment`: Production environment (3 levels) - `weaning_weight_kg`: Weight at weaning - **Structure**: Breeds nested within environments **beef_multifarm_unbalanced.csv**: - `steer_id`: Individual steer identifier - `farm`: Farm/feedlot identifier (5 farms) - `breed`: Beef breed (4 breeds) - `adg`: Average daily gain - `initial_weight_kg`: Weight at feedlot entry - **Structure**: Missing breed × farm combinations, highly unbalanced :::{.callout-important} ## Rank Deficiency Week 12 datasets demonstrate: - **Unequal sample sizes**: Loss of orthogonality - **Missing cells**: Design matrix **X** is not full rank - **Normal equations**: **X'X** is singular, requires generalized inverse - **Estimability**: Individual parameters may not be estimable, but contrasts often are - **Type III SS**: Essential for unbalanced data ::: --- ## Week 13: Special Topics I {#sec-datasets-week13} **Statistical Focus**: Handle constraints, check estimability, preview genetic evaluation concepts (unequal progeny). | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `swine_litter.csv` | Swine | 31 | sow_id, dam_id, litter_size | Litter size | Unequal progeny per dam | | `layer_egg_production.csv` | Poultry | 45 | hen_id, sire_id, eggs_month | Egg production | Sire comparison, unequal daughters | | `beef_feedlot.csv` | Beef | 162 | steer_id, lot, breed, diet, adg, backfat_mm | Multi-trait feedlot | Complex multi-factor design | | `dairy_sire.csv` | Dairy | 95 | daughter_id, sire_id, sire_name, milk_yield | Milk yield | **Genetic evaluation**, unequal daughters (3-25 per sire) | | `broiler_bodyweight.csv` | Poultry | 89 | bird_id, strain, sex, body_weight_kg | Body weight | Multi-strain comparison | **Variable Details**: **dairy_sire.csv**: - `daughter_id`: Individual daughter (cow) identifier - `sire_id`: Sire identifier (10 sires) - `sire_name`: Sire name/code - `milk_yield`: Daughter's milk production in kilograms - **Structure**: Highly unbalanced (3 to 25 daughters per sire) **beef_feedlot.csv**: - `steer_id`: Individual steer identifier - `lot`: Feedlot pen/lot identifier - `breed`: Beef breed (3 breeds) - `diet`: Dietary treatment (4 diets) - `adg`: Average daily gain - `backfat_mm`: Backfat thickness at slaughter :::{.callout-note} ## Genetic Evaluation Preview `dairy_sire.csv` introduces concepts critical for animal breeding: - **Unequal information**: Some sires have many daughters (high reliability), others have few (low reliability) - **Shrinkage**: Sires with few daughters are "shrunk" toward the population mean - **Best Linear Unbiased Prediction (BLUP)**: Preview of Week 14 mixed models - **Estimability**: Sire differences are estimable even though individual sire effects are not ::: --- ## Week 14: Special Topics II {#sec-datasets-week14} **Statistical Focus**: Polynomial regression, weighted least squares, no-intercept models, mixed model preview. | Dataset Name | Species | n | Variables | Trait(s) | Purpose/Features | |--------------|---------|---|-----------|----------|------------------| | `broiler_growth.csv` | Poultry | 6 | age_days, weight_kg | Growth curve | **Reused from Week 4**, polynomial regression | | `lactation_curve.csv` | Dairy | 60 | cow_id, days_in_milk, milk_yield_kg | Lactation curve | 2nd/3rd degree polynomials | | `pen_avg_adg.csv` | Swine | 5 | pen, adg, pen_size | Pen average ADG | Hand-calculable WLS example | | `feed_efficiency_pens.csv` | Swine/Poultry | 30 | pen_id, diet, fcr, pen_size | FCR by pen | WLS, heterogeneous variance | | `milk_feed.csv` | Dairy | 5 | feed_intake_kg, milk_yield_kg | Milk vs feed | Through-origin regression | | `broiler_gain_feed.csv` | Poultry | 40 | bird_id, feed_intake_kg, weight_gain_kg | Feed efficiency | Gain vs feed | | `simple_sire_model.csv` | Dairy | 20 | daughter_id, sire_id, milk_yield | Milk by sire | Mixed model preview, 3-25 daughters | | `dairy_sire_eval.csv` | Dairy | 100 | daughter_id, sire_id, herd, milk_yield_kg | Sire evaluation | Comprehensive mixed model preview | **Variable Details**: **lactation_curve.csv**: - `cow_id`: Unique cow identifier - `days_in_milk`: Days since calving (0-305) - `milk_yield_kg`: Daily milk yield - **Purpose**: Fit polynomial models (quadratic, cubic) to capture lactation curve shape **pen_avg_adg.csv**: - `pen`: Pen identifier - `adg`: Pen average daily gain - `pen_size`: Number of pigs in pen - **Purpose**: WLS with weights = pen_size (variance ∝ 1/n) **milk_feed.csv**: - `feed_intake_kg`: Daily feed consumption - `milk_yield_kg`: Daily milk production - **Purpose**: No-intercept model (biological constraint: zero feed → zero milk) **dairy_sire_eval.csv**: - `daughter_id`: Individual daughter identifier - `sire_id`: Sire identifier (10 sires) - `herd`: Herd identifier (contemporary group) - `milk_yield_kg`: Daughter's milk yield - **Purpose**: Full mixed model with fixed (herd) and random (sire) effects :::{.callout-note} ## Dataset Reuse `broiler_growth.csv` first appeared in Week 4 for simple linear regression. In Week 14, the same data demonstrates: - **Linear model** (Week 4): weight = β₀ + β₁(age) - **Quadratic model** (Week 14): weight = β₀ + β₁(age) + β₂(age²) - **Cubic model** (Week 14): weight = β₀ + β₁(age) + β₂(age²) + β₃(age³) This shows how the same data can be analyzed with different models depending on the research question. ::: --- ## Week 15: Capstone Project {#sec-datasets-week15} **Statistical Focus**: Integrate all concepts from Weeks 1-14 in a comprehensive multi-breed, multi-farm beef cattle analysis. :::{.callout-important} ## Week 15: Dataset Under Development The Week 15 capstone dataset is planned but not yet created. The planned structure: **Dataset**: `beef_capstone.csv` - **Species**: Beef cattle (steers) - **n**: ~120 observations - **Breeds**: Angus, Hereford, Charolais (3 levels) - **Farms**: 5 feedlots (unbalanced design) - **Traits**: - `adg`: Average daily gain (kg/day) - `initial_weight_kg`: Weight at feedlot entry - `days_on_feed`: Duration of feedlot period - **Design Features**: - Missing breed × farm cells (not all breeds on all farms) - Unbalanced sample sizes (4-15 steers per cell) - Intentional outliers for diagnostic practice - Covariate (initial weight) for ANCOVA - Rank deficiency requiring generalized inverse **Analysis Tasks**: 1. Exploratory data analysis and design matrix construction 2. Multiple regression with covariates 3. Two-way ANOVA (breed × farm) with interaction testing 4. Model diagnostics (outliers, leverage, influence) 5. Handle rank deficiency and estimability 6. Build custom least squares solver from scratch 7. Compare manual calculations with `lm()` output Students integrate all course concepts in this comprehensive real-world scenario. ::: --- ## Species Index {#sec-datasets-species} Datasets grouped by livestock species for easy lookup. ### Dairy Cattle (13 datasets) | Dataset | Week | n | Traits | |---------|------|---|--------| | dairy_milk_practice.csv | 1 | 30 | Milk yield | | dairy_lactation.csv | 4 | 30 | Milk vs DIM | | dairy_milk_breeds.csv | 7 | 12 | Milk by breed | | dairy_milk_fat_breed_diet.csv | 9 | 54 | Milk composition, factorial | | dairy_milk_herds.csv | 10 | 40 | Milk by herd, adjust for DIM | | dairy_lactation_variance.csv | 11 | 40 | Heteroscedasticity example | | dairy_scc_mastitis.csv | 11 | 100 | SCC, outliers/influence | | dairy_herds_unbalanced.csv | 12 | 45 | Multi-herd unbalanced | | dairy_sire.csv | 13 | 95 | Sire evaluation, unequal progeny | | lactation_curve.csv | 14 | 60 | Polynomial lactation curve | | milk_feed.csv | 14 | 5 | Through-origin regression | | simple_sire_model.csv | 14 | 20 | Mixed model preview | | dairy_sire_eval.csv | 14 | 100 | Comprehensive sire evaluation | ### Beef Cattle (9 datasets) | Dataset | Week | n | Traits | |---------|------|---|--------| | beef_carcass_marbling.csv | 6 | 50 | Marbling, multiple predictors | | beef_breeds_small.csv | 8 | 12 | Carcass weight, contrasts | | beef_feedlot_adg.csv | 10 | 40 | ADG, ANCOVA | | beef_marbling.csv | 11 | 9 | Marbling with outlier | | beef_feedlot_gain.csv | 11 | 45 | ADG, diagnostics | | beef_multifarm_unbalanced.csv | 12 | 177 | Largest dataset, missing cells | | beef_feedlot.csv | 13 | 162 | Multi-factor feedlot | | beef_capstone.csv (planned) | 15 | ~120 | Comprehensive capstone | ### Swine (8 datasets) | Dataset | Week | n | Traits | |---------|------|---|--------| | pig_litter_breeds.csv | 3 | 6 | Litter size by breed | | pig_feed_efficiency.csv | 4 | 20 | ADG vs feed intake | | litter_size_genetics.csv | 7 | 32 | Litter size by line | | swine_growth_unbalanced.csv | 7 | 29 | ADG unbalanced | | swine_growth_breed_diet.csv | 9 | 8 | Factorial 2×2 | | swine_litter_parity.csv | 10 | 9 | Litter size ANCOVA | | swine_birth_weight_outliers.csv | 11 | 40 | Intentional outliers | | swine_litter.csv | 13 | 31 | Unequal progeny | | pen_avg_adg.csv | 14 | 5 | WLS example | | feed_efficiency_pens.csv | 14 | 30 | WLS heterogeneous variance | ### Poultry (12 datasets) | Dataset | Week | n | Traits | |---------|------|---|--------| | broiler_bodyweight_sex.csv | 3 | 20 | Weight by sex | | broiler_growth.csv | 4, 14 | 4-6 | Growth curve (reused) | | broiler_feed_efficiency.csv | 7 | 50 | FCR by ration | | layer_strains_egg.csv | 8 | 50 | Egg production | | layer_egg_bodyweight.csv | 10 | 30 | Eggs ANCOVA | | broiler_bodyweight_growth.csv | 11 | 50 | Growth diagnostics | | poultry_fcr.csv | 11 | 50 | FCR diagnostics | | layer_egg_production.csv | 13 | 45 | Sire comparison | | broiler_bodyweight.csv | 13 | 89 | Multi-strain | | broiler_gain_feed.csv | 14 | 40 | Feed efficiency | ### Sheep (6 datasets) | Dataset | Week | n | Traits | |---------|------|---|--------| | sheep_weight.csv | 4 | 5 | Birth to weaning weight | | lamb_data.csv | 5 | 40 | Weaning weight prediction | | lamb_growth_breed_diet.csv | 9 | 12 | Factorial growth | | lamb_weaning_weight.csv | 11 | 60 | Comprehensive diagnostics | | sheep_fleece_unbalanced.csv | 12 | 6 | Fleece weight unbalanced | | lamb_breed_environment.csv | 12 | 72 | Nested design | --- ## Dataset Size Guide {#sec-datasets-size} Datasets organized by sample size to help find appropriate examples for different learning goals. ### Hand-Calculable (4-12 observations) Perfect for manual matrix calculations, verifying software output, and understanding mechanics. | Dataset | Week | n | Species | Purpose | |---------|------|---|---------|---------| | broiler_growth.csv | 4 | 4 | Poultry | Simple regression by hand | | sheep_weight.csv | 4 | 5 | Sheep | Regression practice | | milk_feed.csv | 14 | 5 | Dairy | Through-origin | | pen_avg_adg.csv | 14 | 5 | Swine | WLS by hand | | pig_litter_breeds.csv | 3 | 6 | Swine | Design matrix | | sheep_fleece_unbalanced.csv | 12 | 6 | Sheep | Unbalanced by hand | | swine_growth_breed_diet.csv | 9 | 8 | Swine | 2×2 factorial | | beef_marbling.csv | 11 | 9 | Beef | Outlier detection | | swine_litter_parity.csv | 10 | 9 | Swine | ANCOVA by hand | | beef_breeds_small.csv | 8 | 12 | Beef | Contrasts | | dairy_milk_breeds.csv | 7 | 12 | Dairy | One-way ANOVA | | lamb_growth_breed_diet.csv | 9 | 12 | Sheep | Factorial | ### Moderate (20-60 observations) Realistic sample sizes for most livestock experiments, transition from hand calculation to software. | Dataset | Week | n | Species | |---------|------|---|---------| | pig_feed_efficiency.csv | 4 | 20 | Swine | | broiler_bodyweight_sex.csv | 3 | 20 | Poultry | | simple_sire_model.csv | 14 | 20 | Dairy | | swine_growth_unbalanced.csv | 7 | 29 | Swine | | dairy_milk_practice.csv | 1 | 30 | Dairy | | dairy_lactation.csv | 4 | 30 | Dairy | | layer_egg_bodyweight.csv | 10 | 30 | Poultry | | feed_efficiency_pens.csv | 14 | 30 | Swine | | swine_litter.csv | 13 | 31 | Swine | | litter_size_genetics.csv | 7 | 32 | Swine | | lamb_data.csv | 5 | 40 | Sheep | | beef_feedlot_adg.csv | 10 | 40 | Beef | | dairy_milk_herds.csv | 10 | 40 | Dairy | | dairy_lactation_variance.csv | 11 | 40 | Dairy | | swine_birth_weight_outliers.csv | 11 | 40 | Swine | | broiler_gain_feed.csv | 14 | 40 | Poultry | | dairy_herds_unbalanced.csv | 12 | 45 | Dairy | | beef_feedlot_gain.csv | 11 | 45 | Beef | | layer_egg_production.csv | 13 | 45 | Poultry | | beef_carcass_marbling.csv | 6 | 50 | Beef | | broiler_feed_efficiency.csv | 7 | 50 | Poultry | | layer_strains_egg.csv | 8 | 50 | Poultry | | broiler_bodyweight_growth.csv | 11 | 50 | Poultry | | poultry_fcr.csv | 11 | 50 | Poultry | | dairy_milk_fat_breed_diet.csv | 9 | 54 | Dairy | | lactation_curve.csv | 14 | 60 | Dairy | | lamb_weaning_weight.csv | 11 | 60 | Sheep | ### Large (60+ observations) Complex datasets for advanced analyses, diagnostics, and realistic production scenarios. | Dataset | Week | n | Species | Features | |---------|------|---|---------|----------| | lamb_breed_environment.csv | 12 | 72 | Sheep | Nested design | | broiler_bodyweight.csv | 13 | 89 | Poultry | Multi-strain | | dairy_sire.csv | 13 | 95 | Dairy | Genetic evaluation | | dairy_scc_mastitis.csv | 11 | 100 | Dairy | Outliers/influence | | dairy_sire_eval.csv | 14 | 100 | Dairy | Mixed model | | beef_feedlot.csv | 13 | 162 | Beef | Multi-factor complex | | beef_multifarm_unbalanced.csv | 12 | 177 | Beef | Largest, missing cells | --- ## Quick Reference: Alphabetical Listing {#sec-datasets-quick-ref} All 48 datasets in alphabetical order with key information. | Dataset | Week | Species | n | Primary Trait(s) | |---------|------|---------|---|------------------| | beef_breeds_small.csv | 8 | Beef | 12 | Carcass weight | | beef_capstone.csv (planned) | 15 | Beef | ~120 | ADG multi-factor | | beef_carcass_marbling.csv | 6 | Beef | 50 | Marbling quality | | beef_feedlot.csv | 13 | Beef | 162 | Multi-trait feedlot | | beef_feedlot_adg.csv | 10 | Beef | 40 | ADG ANCOVA | | beef_feedlot_gain.csv | 11 | Beef | 45 | ADG diagnostics | | beef_marbling.csv | 11 | Beef | 9 | Marbling with outlier | | beef_multifarm_unbalanced.csv | 12 | Beef | 177 | ADG unbalanced | | broiler_bodyweight.csv | 13 | Poultry | 89 | Body weight multi-strain | | broiler_bodyweight_growth.csv | 11 | Poultry | 50 | Growth diagnostics | | broiler_bodyweight_sex.csv | 3 | Poultry | 20 | Weight by sex | | broiler_feed_efficiency.csv | 7 | Poultry | 50 | FCR by ration | | broiler_gain_feed.csv | 14 | Poultry | 40 | Feed efficiency | | broiler_growth.csv | 4, 14 | Poultry | 4-6 | Growth curve | | dairy_herds_unbalanced.csv | 12 | Dairy | 45 | Milk unbalanced | | dairy_lactation.csv | 4 | Dairy | 30 | Lactation curve | | dairy_lactation_variance.csv | 11 | Dairy | 40 | Heteroscedasticity | | dairy_milk_breeds.csv | 7 | Dairy | 12 | Milk by breed | | dairy_milk_fat_breed_diet.csv | 9 | Dairy | 54 | Milk composition | | dairy_milk_herds.csv | 10 | Dairy | 40 | Milk by herd | | dairy_milk_practice.csv | 1 | Dairy | 30 | Milk yield | | dairy_scc_mastitis.csv | 11 | Dairy | 100 | SCC outliers | | dairy_sire.csv | 13 | Dairy | 95 | Sire evaluation | | dairy_sire_eval.csv | 14 | Dairy | 100 | Mixed model sire | | feed_efficiency_pens.csv | 14 | Swine | 30 | FCR WLS | | lactation_curve.csv | 14 | Dairy | 60 | Polynomial lactation | | lamb_breed_environment.csv | 12 | Sheep | 72 | Nested design | | lamb_data.csv | 5 | Sheep | 40 | Weaning weight | | lamb_growth_breed_diet.csv | 9 | Sheep | 12 | Factorial growth | | lamb_weaning_weight.csv | 11 | Sheep | 60 | Diagnostics | | layer_egg_bodyweight.csv | 10 | Poultry | 30 | Eggs ANCOVA | | layer_egg_production.csv | 13 | Poultry | 45 | Sire comparison | | layer_strains_egg.csv | 8 | Poultry | 50 | Egg production | | litter_size_genetics.csv | 7 | Swine | 32 | Litter size | | milk_feed.csv | 14 | Dairy | 5 | Through-origin | | pen_avg_adg.csv | 14 | Swine | 5 | WLS pens | | pig_feed_efficiency.csv | 4 | Swine | 20 | ADG vs feed | | pig_litter_breeds.csv | 3 | Swine | 6 | Litter size | | poultry_fcr.csv | 11 | Poultry | 50 | FCR diagnostics | | sheep_fleece_unbalanced.csv | 12 | Sheep | 6 | Fleece weight | | sheep_weight.csv | 4 | Sheep | 5 | Birth to weaning | | simple_sire_model.csv | 14 | Dairy | 20 | Mixed model preview | | swine_birth_weight_outliers.csv | 11 | Swine | 40 | Outliers | | swine_growth_breed_diet.csv | 9 | Swine | 8 | Factorial 2×2 | | swine_growth_unbalanced.csv | 7 | Swine | 29 | ADG unbalanced | | swine_litter.csv | 13 | Swine | 31 | Unequal progeny | | swine_litter_parity.csv | 10 | Swine | 9 | Litter ANCOVA | --- ## Notes and Future Plans {#sec-datasets-notes} ### Missing Data Conventions - Missing values are coded as `NA` in all CSV files - Students should check for missing data before analysis - Missing data is rare in these datasets (most are complete cases) - When present, missingness is documented in variable descriptions ### Data Integrity All datasets have been verified for: - Correct variable names matching descriptions - Realistic trait values for each species - Appropriate sample sizes for pedagogical purposes - Consistent units (kg, days, etc.) ### Accessing Data Files All datasets are available in week-specific subdirectories: - Path pattern: `WeekXX_TopicName/data/dataset_name.csv` - Files are plain-text CSV with Unix line endings - First row contains variable names (headers) - No row names included ### Future Dataset Development **Week 15 Capstone Dataset**: - Currently under development - Will provide comprehensive integration of all course concepts - Designed to challenge students with realistic complexity - Includes all major features: unbalanced design, covariates, outliers, rank deficiency **Potential Additions**: - Additional aquaculture datasets (tilapia, salmon) - Horse breeding examples - Multi-trait genetic evaluation datasets - Longitudinal/repeated measures examples ### Citation and Use These datasets were created specifically for educational purposes in this linear models course. While inspired by real livestock production systems, all data values are simulated to ensure appropriate pedagogical features. Students may use these datasets for practice and learning but should cite this course if using them in reports or presentations. ### Feedback If you encounter issues with any dataset (incorrect values, unclear variables, missing files), please report to the course instructor. Continuous improvement of these educational resources is a priority. --- This completes Appendix D. For mathematical notation, see [Appendix A](#sec-appendix-a). For R function reference, see [Appendix C](#sec-appendix-c).