I'm doing a literature review, and most of the data I need come from linear regression analyses. There's a paper that has solid data, but it doesn't report r-squared, which is the metric I need. It reports sample size, beta, standard error, and the p-value of the model. Is it possible for me to calculate an overall r-squared value from those?
If it helps, the source I'm looking at is Byun & Tiede (2017) in PLOS One: "Perception-production relations in later development of American English rhotics," and it's pages 13 through 15.
Thanks!
No. It is not possible to calculate the r-squared value from these pieces of information. At least you need information on the residual standard error or the correlation between y and each of the predictors. See here: http://phd.meghan-smith.com/wp-content/uploads/2016/05/MR14-57-70-Intro-to-MR.pdf
It is possible only in the case of the regression model with 1 predictor, as ß, in this case, is equal to Pearson's r, which is directly related to R-squared.
The coefficients reported in the paper are from mixed models and from multiple regression models. I don't see a way to get the model R² here. It might be easiest if you just contact the authors (Tara Byon, [email protected]) and ask for the R² values or for the data so that you could calculate the R² values yourself.
- Badges
- Science topic
- Similar topics
- Linguistics
- Philology
- Language
- English