spssforum.com

Hello

I found some very bizar results, so I would like to check how this is possible.

First analysis:
I did a number of linear regression analyses. In a first block, I added my three control variables. In the second block, I added my four independent testing variables to the model, so a total of seven variables.

Second analysis:
Then I did the same again, but instead of adding the four variables in the second block, I only added one of these four (because of multicollinearity with the other three independents). So the first block has again the three control variables, my second block has the three control variables and one independent that is tested.

The question:
As the first block of the first analysis is exactly the same as the first block of the second analysis, this should give identical results, isn't it? Well, it isn't! How is that possible? Checking further, it turns out that only analyzing the first block (so not adding a second block) gives the results of the second analysis. Why is the result in the first analysis different? Is this an error is SPSS?

You can see the results in attachment.

Thank you for your help!

Because the sample sizes are not the same.

In your analysis 1, the df2 is 52; but in analysis 2, the df2 is 55. Aka, there three more subjects included in analysis 2.

The reason is probably due to missing data in independent 2, 3, and 4 in analysis 1. I bet there are three cases in analysis 1 that had missings in independent 2, 3, and 4, and hence they were excluded from the overall analysis, lowering the df2 of block 1 in analysis 1 by 3.

Wow, super, thank you very much for this quick replay and for this solution. That is indeed the answer!