Some words on Lab 4 + help with coefficients

Nicky Wakim

2024-03-18

Lab 4

Overall

• Interactions: Limit to 1-2 interactions!!

  • With large sample size, many will come up significant
  • Make choice based on your visualizations and previous research

• When deciding on what covariates to leave in or take out of a model:

  • If the covariate is categorical with more than 2 levels, you MUST use F-test!!

• Check you HTML document in the report!!

  • Make sure things look good: proportions and font size should make sense
  • Tables and figures should look good!

• When looking at % change:

  • Some only looked at model with explanatory variable and variable in question

  • Need to use the model with all other variables as well!!

Project Report

Table 1

Good page on creating a Table 1

Displaying coefficients

Helpful source for plotting predictions for variables that are part of interactions!

• I would make a table or forest plot for the main effects of my model

  • Then make a plot showing the interactions

• Let’s look at it together?

library(ggeffects)
dat <- predict_response(final_model, terms = c("identfat", "comptomost"))
plot(dat)

See recording for results!