Beta coefficients in regression?
Hi
One question!
in JASP when running a regression only standardised beta coefficients are displayed for scale variables.You can of course force categorical variables to scale variables. So, in a multiple regression with both continuous and categorical variables is it possible to interpret the standardised beta coefficients? Thus, can the standardised beta be used to understand which variables (be it continuous or categorical) contribute most to the model?
Grateful for a reply.
All the best Per “the JASP lover”
Comments
See my comment on another thread here: https://forum.cogsci.nl/discussion/comment/27658#Comment_27658
The answer to your question is probably no, it's difficult to compare continuous and categorical predictors together
To the extent that the categorical variable "looks like" a scale variable, this will work. For instance, scores on an IQ test are not really continuous (they are integers), but we always treat them as such. The main problem is with nominal variables. These demand a different treatment, similar to ANCOVA. This is an interesting problem that still requires development.
E.J.
@patc3 and @EJ
Of course if X is a categorical variable, then its standardized coefficient cannot be interpreted as it doesn’t make sense to change X by 1 standard deviation. However, i have heard people saying that in general, this is not a problem since these coefficients are not meant to be interpreted individually, but to be compared to one another in order to get a sense of the importance of each variable in the linear regression model.
But the prior structure differs depending on the scale (continuous vs nominal)
EJ