Regression Recap Notes

[jmgirard]

This is a place to store notes and discuss about the chapter on regression recap

[ammsimmons]

• install -> install.packages (link to helpful documents)

• callout for why each package is necessary (one sentence each,

• why we use tidyverse for easy of analysis, easystats make tables output the models, ggeffects to plot the models)

• call out: read_csv comes from tidyverse

• explain the formula as used in the syntax of lm().

• separate MR into three models, two continous var, one dummy coded, and one with continous and dummy code.
  -- categorial/dummy coding : means and contrasts
  -- with continous and dummy coded: "we controlled"

• before scatterplot, put histogram for both

• ggplot also comes from tidyverse (use as a footnote)

• tel them what model_paramters() is

• explain the 95% confidence interval

• one sd in yearas is 0.5 increase sd in salary.

• explain using pipes (in introduction).

[ammsimmons]

New notes:

• for packages, use the website not the pdfs.
• link to how to use ggplot(), read this chapter for R for Data Science, or the ggplot2() book.
• move lm() syntax callout before the fititng code block.
• put 1 + predictor for the lm() syntax block
• uncomment out the standardized_paramters(fit)
• reread for typos.
• adjusted R^2 typo.
• remove "inference" parens
• mytheme invisible code chunk, and add "+mytheme" to every plot, and state in Intro page.
• what does it mean to control for multple predictors, and why you wanna do that? (one sentence. For example:... "unique contribution of each predictor")... repeat what I had a little further down.
• remove "and" in parameter table and interpretation
• use the term "slopes", but in parens say (beta1 and beta2).
• typo, subscripting the X2 and X2 (section 4.1)
• Why would we prefer zero order vs partial effects (decision you need to make in conjunction with your research question, they answer different questions etc).
• "the unique effect when controlling for other predictors explicitly in the model and can't say anything when controlling for variables not in the model".
• "As a reminder, the $ extracts one of the columns in the dataframe" 4.3
• print out yearspubs (2.2)
• move variable table to after loading it and printing yearspub (2.2).
• equation for 4.2 categorical, C(i) needs to be subscripted
• 4.2.1, explicltly put (by = "sex)
• 4.2: "Note that if you have two more categories, you will use the exact same approach but have more than one slope".
• 4.3.1: specifically, these means for males and females "averaged over" not "across".
  -4.4: beta symbol is not formatted properly.
• 4.4: moderation is an umnbella term, and interaction is one type. SO JUST USE THE TERM "INTERACTION"... and say (which is a form of moderation). Also, say interaction terms is "multiplying the predictors together".
• 4.4: Use the long form of interaction syntax. then callout for asterisk.
• use print_md for 4.4 output of marginal effects.
• 4.4: callout, the interaction was not significnt, but one of the simple slopes was and the other wasnt.
• plots for assumptions: height 7, width 5 ... fig-dash ?

[ammsimmons]

Still need to change the plot of assumptions with fig-dash??!