1
$\begingroup$

I am running a hierarchical regression model. (That is, I am running a series of regression models, adding additional variables each time so that each model is nested within the next. Here, each model is called a "block".) I have four ordinal Likert-style independent variables (named IV1, IV2, IV3, IV4, all centered) and one dependent variable (named DV).

However, I am interested only in selected hypotheses / relationships which are:

Direct effect

  • H1: IV1 >>> DV

2-way interactions (moderations)

  • H2: IV1 >>> DV moderated by IV2
  • H3: IV1 >>> DV moderated by IV3

3-way interactions (moderated moderation)

  • H4: IV1 >>> DV moderated moderation with IV2 * IV4
  • H5: IV1 >>> DV moderated moderation with IV3 * IV4

I am conducting a hierarchical regression using SPSS.

  • Block 1: only controls
  • Block 2: direct effects (IV1, IV2, IV3, IV4)
  • Block 3: 2-way interactions Block 4: 3-way interactions

Questions:

  1. Regarding Block 2 and 3: do I need to include all possible 2-way / 3-way interactions or only the ones which I am hypothesizing?
  2. From which block do I read the results? For example, for the direct effects would I look only at block 2 or the "full model" from block 4?
$\endgroup$
  • $\begingroup$ This appears to be essentially asking for Stata guidance, but without specifying anything about the commands to be used. As such it's off-topic here. See the Help Center for guidance on software-related questions. In any forum supporting Stata questions showing code so far and being precise about data structure and variable names would be basic expectations. $\endgroup$ – Nick Cox Jul 8 '16 at 7:09
  • $\begingroup$ Dear Nick, thanks for your quick answer. I am sorry if I have not been specific enough. I really do not need any stata specific guidance or commands. More details on the data structure: sample n > 500, all items are ordinal (likert-style), some of the controls are binary (dummy variables), no time series but collected from a survey. I have further included the command in the original post now if that helps. I have named the variables as noted in the original post. The questions are not Stata specific but concern the statistical build-up and, I guess, interpretation of the results. Thanks! $\endgroup$ – Christoph C Holt Jul 8 '16 at 7:25
  • $\begingroup$ Whatever that syntax is, it is not Stata syntax!!! SPSS? I am afraid you can't easily it have both ways. If what you are doing requires understanding of some syntax, it is not generically posed. "From which block do I read the results" is hard to understand except in software terms. $\endgroup$ – Nick Cox Jul 8 '16 at 8:20
  • $\begingroup$ Oh boy, I think I have had a complete melt-down this morning. Obviously you are right, and I was wrong - I am using SPSS. Yet, I believe the questions are valid regardless of the program: 1) Is it necessary to include all possible variations of 2-way and 3-way interactions in the calculation even if one only wants to test certain hypotheses regarding 2-way and 3-way interactions? 2) Also regardlessof the statistical program, in the reported data as published in papers, would one, e.g., for the direct effects, reference to the first block which only includes direct effects or to the last block $\endgroup$ – Christoph C Holt Jul 8 '16 at 8:32
  • $\begingroup$ Welcome to the site, @ChristophCHolt. I don't believe you are actually asking for SPSS code, but have an on-topic statistical question. I edited your post to make that clearer. Please check to ensure that the post reflects what you want to ask. If it doesn't, please roll back the edits with my apologies. $\endgroup$ – gung Jul 8 '16 at 17:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.