Comprehensive Linear Modeling with R

Explore how to infer room width using rcmdr by loading data, converting units to a consistent basis, and comparing mean estimates with grouped numerical summaries and box plots.

Subset data to create an active data set in R Commander and compare groups using Welch and Wilcoxon tests. Evaluate normality with Q-Q plots and Shapiro-Wilk tests to confirm assumptions.

Analyze the waves data by comparing two morphing methods using a difference variable, box plots, q-q plots, and a paired t-test, with an optional Wilcoxon nonparametric check.

Explore non-parametric inference with paired Wilcoxon tests, and assess relationships between water hardness and mortality across north and south towns using box plots, scatter plots, and Pearson correlation.

Revisit conditional inference using a permutation test to reanalyze a room width study, converting feet and meters to a common unit and comparing mean estimates.

Compare resident and physician counts of birth anomalies using a matrix of paired observations, apply a marginal-homogeneity test from coin, and interpret significant differences in marginal estimates.

Practice drawing a variety of plots by manipulating raw variables with arithmetic in chapter 2. Learn to label scatterplot points using the text function via help tips.

Examine graphical analysis of household spending by category (housing, food, goods and service) and gender, compute proportions from total expenditures, and compare men and women using scatterplots with regression lines.

Learn to create multiple plots in R: five side-by-side box plots of suicide rates by age groups, a labeled scatterplot matrix, and a trellised plot of homicide by income.

Explore conditional inference by creating two normal groups with different means and comparing them with exact, Wilcoxon rank-sum, and permutation tests, using a script and simulations.

The lecture demonstrates creating a four-level interaction variable in Rcmdr by combining source and type, then analyzing weight gain in rats to reveal a potential beef x protein interaction.

Learn to model weight gain and foster feeding in rats with R, using lm and ANOVA to analyze main effects and interactions in unbalanced designs.

Apply a multivariate analysis of variance to test if north versus south location affects water hardness and mortality, using two dependent variables and Wilks' test.

Explores how five time-period groups of skull measurements from ancient egypt vary over time to test evolution versus interbreeding. Shows how to summarize by time period and explore the relationships among skull measurements like bh and bl through plots, with a dataset of 150 observations.

Explore relationships among four skull measurements with a scatterplot matrix of numerical variables, then run a MANOVA across five time periods and interpret pairwise time-period differences.

Engage with two exercise sets in your folder, focusing on inference, with one set a bit harder and a couple of simple problems about a Nova.

Explore estimating the age of the universe through a simple regression of galaxies' velocities on their distances using linear modeling in R with the Hubble data.

Explore testing hypotheses and estimating uncertainty in linear models with R, using t-statistics, p-values, confidence intervals, and AIC to compare models and interpret beta coefficients.

Explore how cloud seeding with silver iodide interacts with cloud tops and prior rainfall to affect rainfall, using a suitability criterion and 24 days of Florida data.

Explore rainfall relationships through scatterplot matrix, identify outliers, and fit a linear model in R to test seeding interactions with ecological factors, highlighting significant interaction effects.

Use standard diagnostic plots to evaluate fitted models, compare M1 and M2, and note how removing two outliers (rows 1 and 15) shifts seeding interactions with wetness and other factors.

Assess model assumptions with error checking plots, examine residuals, q-q plots, and Cookes distance to evaluate normality and leverage in linear modeling.

Examine how extreme x-values and outliers affect model fit and slope. Assess leverage and Cook's distance, then explore survival plots and Cox regression assumptions.

Explore how plasma proteins relate to the erythrocyte sedimentation rate using a binary generalized linear model with binomial error and logit link, focusing on fibrinogen and globulin and odds ratios.

Apply logistic regression in a generalized linear model to predict ESR rate as a binary outcome using fibrinogen and globulin; compare models with AIC and deviance for fit.

Compare two models predicting ESR with an analysis of variance; when chi-squared is not significant, choose the simpler model per Ockham's razor, noting fibrinogen's odds ratio for unhealthy ESR.

Explore how education and gender influence agreement with the statement on women's roles, using a 42-row factorial design of counts and a binomial glm with a logit link.

Explore how education relates to the probability of agreeing with a statement using R, computing predicted values, plotting by gender, and testing an education by gender interaction.

Examine how an interaction term alters gendered education effects on agreement and display results with male/female symbols; evaluate model fit via diagnostic plots, residuals, and Cookes distance.

Evaluate a placebo-controlled trial of colonic polyps using a Poisson GLM with treatment and age as covariate, then address overdispersion with a quasi-Poisson model and interpret residual deviance.

Analyzes whether driving is a risk factor for back pain from acute herniated lumbar discs using a case-control design and clogit in R to estimate odds ratios.

Explore partial effects of six covariates on sulfur dioxide. Use a GAM boosting model with smooth fits to reveal relationships among temperature, population, manufacturing, wind, and precipitation.

Fit a logistic generalized additive model using the GAM framework, estimate smooth terms with cubic splines, plot and interpret partial effects and 95% confidence intervals for kyphosis recurrence.

Explore lowess smoothing for nonparametric, locally weighted regression of population changes, fit and predict with the model, and compare with a decision-tree regression to identify a potential break point.

Explore using anova and t apply to delta levels, filter out missing data, compare north and south means relative to the threshold, and reveal a stairstep relationship.

Model a binary outcome with logistic GAMs in R, using smooth area and isolation terms to predict bird presence on islands, and compare linear versus non-linear fits via deviance.

Explore generalized additive models with the mgcv package, using ozone data to compare nonparametric smooths, pairs plots, and custom panel functions for understanding relationships among wind, temperature, radiation, and ozone.

Explore generalized additive models with the mgcv package, comparing parametric and non-parametric smooth terms and using nested f-tests to select the best model for ozone data.

Explore generalized additive models with the mgcv package, including building interaction terms, comparing models by significance and predictive power, and diagnosing fit with residual plots and prediction.

Learn to build generalized additive models with parametric and non-parametric components, smooth interactions via conditioning plots, and a linear compression ratio, and assess fit via residuals and deviance explained.

Compare two depression treatments, treatment as usual and beat the blues, in 100 participants with five repeated BDI scores over eight months, reshaping data to tall form for covariate analysis.

Learn data wrangling in R for the beat the blues study: reattach covariates, create new columns, convert time to an ordered factor, reorder levels, and create treatment-specific box plots.

Load the lme function from the lme4 package, specify fixed and random effects for a random intercept model, and compare models with analysis of deviance to gauge improvement.

Explore generalized estimating equations for longitudinal data, compare identity, exchangeable, auto regressive, and unstructured correlation matrices, and distinguish marginal from conditional models.

Prepare longitudinal data for GEE analysis by ordering records by subject, then compare two models with independent and exchangeable correlations using Gaussian family, including pretreatment and treatment covariates.

Explains running gee with beat the blues as binomial glm, comparing naive and robust standard errors under different correlation structures, and interpreting coefficient significance and correlation matrices.

Prepare respiratory data in R for a binary outcome, clean month zero records, create a baseline covariate, merge datasets, and code status for logistic regression.

Explore binary outcome modeling for respiratory illness with binomial GLM and GEE in R, assess overdispersion, and interpret coefficients as odds ratios with robust standard errors and confidence intervals.

Explore poisson regression for seizure count data to compare placebo and progabide treatment, including baseline seizures and period as covariates.

Use box plots and log-transform seizure counts with an offset to stabilize variance, then compare GEEs with independence and exchangeable correlations to examine treatment effects under overdispersion.