Hierarchical regression JASP using a mixture of entry methods

RachelKing42

Hello All,

We are currently transitioning from SPSS to JASP for our undergraduate provision (Psychology).

When teaching regression we used to be able to use a mixture of entry methods when running hierarchical regression. With the first block forced entry and then second block using a stepwise entry method.

Is it possible in JASP?

With many thanks! :)

Rachel

patc3

After some testing, I think you can reproduce this using these steps:

1) add all your variables (first and second block) as your predictors

2) under "Model", enter the variables from your first block in the Model 0 box (careful, JASP might add interaction terms automatically, e.g. Jobsat1*Jobsat2, remove those if you didn't mean to add them); no need to touch Model 1

3) Change the method to Stepwise

This seems to reproduce SPSS (even the sequence of models seems to be the same, at least when I tested it)

TarandeepKang

Okay, in principle this might work for this particular case/pair. But in SPSS you can (if I recall) select an entry method separately for each block. You might want a combination not available with this current pairing, and this certainly doesn't extend to models with more than two blocks. I assume therefore that for a more flexible implementation covering a wider range of use cases, we might need to ask the programmers to explicitly move the dropdown menu so that a different choice is possible per block? What do you think @patc3 and @RachelKing42?

patc3

@TarandeepKang hi Tarandeep, to me this doesn't seem like time well invested, I've personally never actually used hierarchical regression outside of classes, and I do this for a living. Just my 2¢

TarandeepKang

@patc3 Going to assume that you later in your career than me, given that I'm only a PhD student! I'm also going to say that the use of hierarchical regression in real research situations is somewhat dependent, even likely dependent, on subfield of research. I'll say also that it seems to be prevalent in my field, health psychology and behavioural medicine, and in adjacent areas of social psychology. I could point you in a couple of publications that use this, but frankly, I find that embarrassing since we are only talking about very elementary statistics! :-)I would however agree, that this particular use case of different entry methods for block is so unusual that I've never seen it out in the wild, or in class. So I guess the things on how of a need @RachelKing42 sees it.

PS, I am aware of the new research suggesting we should be using ML as it outperforms multiple regression-based methods anyway, so will try to adopt those in future!

Fife, D.A., D’Onofrio, J. Common, uncommon, and novel applications of random forest in psychological research. Behav Res 55, 2447–2466 (2023). https://doi.org/10.3758/s13428-022-01901-9