SPSS Statistics

I have run a series of stepwise regression models for a transit authority, in hopes of identifying priority areas for improvement. Many of my independent variables are highly correlated with one another. Is there a subcommand in regression that will allow me to take into account interaction effects, or perhaps a way to import interaction coefficients from other procedures?

Thanks in advance

Bob Walker

I have run a series of stepwise regression models for a transit authority, in hopes of identifying priority areas for improvement. Many of my independent variables are highly correlated with one another. Is there a subcommand in regression that will allow me to take into account interaction effects, or perhaps a way to import interaction coefficients from other procedures?

Thanks in advance

Bob Walker

I support Jon's comments about the perils of "stepwise" variable selection methods.  More info about those perils can be found here:

• https://www.stata.com/support/faqs/statistics/stepwise-regression-problems/
• https://discourse.datamethods.org/t/author-checklist/3407#arrow_rightuse-of-stepwise-variable-selection-24

I hope this helps.

I have run a series of stepwise regression models for a transit authority, in hopes of identifying priority areas for improvement. Many of my independent variables are highly correlated with one another. Is there a subcommand in regression that will allow me to take into account interaction effects, or perhaps a way to import interaction coefficients from other procedures?

Thanks in advance

Bob Walker

I have run a series of stepwise regression models for a transit authority, in hopes of identifying priority areas for improvement. Many of my independent variables are highly correlated with one another. Is there a subcommand in regression that will allow me to take into account interaction effects, or perhaps a way to import interaction coefficients from other procedures?

Thanks in advance

Bob Walker

Thank you Jon, super helpful! The GLM Univariate approach with full factorial option for the covariates worked nicely to capture interaction effects, which (compared to stepwise without them) were minimal. I especially liked the STAT RELIMP procedure, because I can scale the Shapely values to sum to 100% and get relative contributions. From what I understand, Shapely values capture interactions and dependencies that regression weights do not? Regardless, this approach will be easier for the client to digest.

Thanks again,

Bob Walker

I have run a series of stepwise regression models for a transit authority, in hopes of identifying priority areas for improvement. Many of my independent variables are highly correlated with one another. Is there a subcommand in regression that will allow me to take into account interaction effects, or perhaps a way to import interaction coefficients from other procedures?

Thanks in advance

Bob Walker