How to Control for Variables in JASP in Multiple Linear Regression

saseeri

Hello everyone,

I'm currently working on a multiple linear regression analysis in JASP and I need some guidance on how to properly control for certain variables. Specifically, I want to understand the best way to enter control variables using JASP.

Here’s my situation:

1. I have a set of 5 control variables that I need to account for and one independent categorical variable on my dependent variable.

My specific questions are:

• How do I enter control variables and independent variables in the correct order in JASP for multiple linear regression?
• Is it necessary to use hierarchical regression, or can I simply enter all variables together and still properly account for the control variables?
• How to do a post hoc test for Multiple Linear Regression for a categorical variable?

Any detailed instructions or best practices on setting this up in JASP would be greatly appreciated!

Thank you in advance for your help!

Regards,

Sahar

EJ

The upcoming version of JASP will have hierarchical regression implemented in a more intuitive way. I will ask a team member to elaborate.

Cheers,

E.J.

JohnnyB

Hi @saseeri ,

To control for certain variables, and see if your variable of interest still adds predictive value on top of the control variables, you can do hierarchical regression, where you add the control variables to the lowest model (commonly called the null model), and then see if the alternative model still performs better than the null model. In JASP, you can do so in the Model tab. Below is an example with a big 5 data set, where we could be interested in the relationship between neuroticism and conscientiousness, while controlling for the other big 5 items (so we add those the null model). Then, we can see if there is still a meaningful improvement of the alternative models' fit (quantified by R-squared, for instance).

The next version of JASP will allow specifying more than 2 models in this way, to allow iterative hierarchical regression.

To do posthoc tests, I would recommend using ANCOVA instead (which is essentially the same linear model), where you add your continuous predictors as covariates. In JASP, the ANCOVA has a posthoc option that you can then use.

Please let me know if anything is still unclear!

Cheers,

Johnny