To do for our new lm function

[arnoudplantinga]

• [x] Add standardized coefficients
• [x] Add effect size estimates for the coefficients
  • [x] Eta squared
  • [x] Partial eta squared
• [x] Find an awesome name (awesome_lm?)

[WillemSleegers]

I checked out effect sizes for regression, and eta squared and partial eta squared don't really fit with regression (they do with GLM).

In fact, I think b and beta coefficients are sufficient estimates of an effect size.

We could, however, consider adding Cohen's f.

[mjbsp]

GLM and regression are essentially the same thing. But, regardless, for standard regression standardized slopes and r^2 of model would be pretty typical of most effect size reporting.

[WillemSleegers]

True, but at least in SPSS the default effect size output for GLMs is eta squared; while it is not for regression. I'm actually not that familiar with glm() in R, but I suppose it works the same as lm(), which means also there eta-squared won't fit very well. There should probably be an eta_squared() function, accepting regression output, that produces those effect sizes.

Anyhow, standardized regression coefficients have been added, and r^2 is already part of the output. =)

[WillemSleegers]

Going for awesome_lm() (I really cannot come up with anything better), and I'm forgetting about the extra effect sizes. They probably fit something else better (e.g., anova()).