1

I am analyzing my 2x3 between subject design with MANOVA in SPSS at the moment. In the main model, there is 2 IV's and 2 DV's. Result of the test is 1 main effect and no interaction effect. In an extended model, I am adding gender as a IV. Results now show interaction effect of gender with both IV's. Hence, I am interpreting those results (the effect of the IV's depends on gender).

Now, I was splitting the file by gender to get a deeper understanding and the results are completely different. Does it make sense to split the file by gender and run manovas again? I am a little puzzled now which results to interpret? Should I split the file or can you recommend any other follow-up tests?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Meryam Folen
  • 11
  • 1
  • 1
    By splitting the file you cannot analyze interactions. In what sense could that possibly provide a "deeper understanding"? What are you hoping to accomplish, specifically? – whuber Jan 12 '19 at 20:37
  • Yes, true.. I saw other analysis that considered gender splitting the file. Just wanted to make sure that its the right procedure. On another note: If the interaction is only significant on 1 DV, do I then interpret the main effect on the second DV? – Meryam Folen Jan 14 '19 at 13:40

2 Answers2

2

Splitting the dataset on gender is not a good idea. You lose the possibility to model and interpret the interactions (and you lose statistical power.) More concretely, splitting the file on gender is the same as including all interactions with gender in the model (well, in addition it gets more variance estimates.) The full details can be found in my answer here: Separate Models vs Flags in the same model

mkt
  • 11,770
  • 9
  • 51
  • 125
kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
1

No, splitting the dataset into two does not gain you anything over simply using the full dataset and modelling the interaction.

mkt
  • 11,770
  • 9
  • 51
  • 125