SPSS - how to save variables for T-tests

SDN Members don't see this ad. (About Ads)

I am wondering if there is any way I can save all the pairs of variables I am comparing?

I have quite a lot to compare and everytime I open SPSS I need to re-select all the different variables.

Thanks in advance

You're running the analysis multiple times, I assume? If you only need to run it once, saving the output is your best bet. (When it gives you the output, which will be in a different window from your data file, click file and save. You should be able to save the .spo file to open later with SPSS).

If you really are running the same or similar analyses over and over (e.g., with slightly different data sets, etc), then the best way to save this analysis is to use syntax. Instead of clicking OK before you run your t-tests, click 'Paste'. This will open a new window with the code that runs the t-test. Save that syntax, and you can open and run it any time in the future.

I hope this helps!

Yep, saving the syntax would've been my suggestion as well. You can then just manually type in/change the variables as/if needed.

Thanks for this, it helped greatly. Yes, I was running the same analysis for different data sets. I was having trouble deciding on which outliers to exclude.

What is the general consensus on outliers (is it excluding anything 3 STD outside dataset)?

Originally Posted by tkuhug

Thanks for this, it helped greatly. Yes, I was running the same analysis for different data sets. I was having trouble deciding on which outliers to exclude.

What is the general consensus on outliers (is it excluding anything 3 STD outside dataset)?

Personally, I usually start with z=3.29 as my cut-off, although up to z=4 might be appropriate for a larger N.

Originally Posted by tkuhug

Thanks for this, it helped greatly. Yes, I was running the same analysis for different data sets. I was having trouble deciding on which outliers to exclude.

What is the general consensus on outliers (is it excluding anything 3 STD outside dataset)?

I would check the Cook's D and Leverage values to see how influential the outliers are. If they aren't influential I would leave them in.

One of the problems with outliers is there is no general consensus. I've typically done +- 3 SDs, but it varies based on a number of factors. Transformations can pull them in a bit, but are generally only done if you also have skew. There are even arguments that unless you have reason to believe they are illegitimate values (i.e. mistakes filling out the form, etc.) that you shouldn't exclude them.

The lack of standardization has been one of the most infuriating things for me in research, especially since I have happened to be involved in multiple studies where the outcomes depended entirely on relatively arbitrary decisions like these. Personally, I'm a fan of "run it several ways, pick one, report one and footnote the others" approach. Reviewers have been mixed in their feelings on it, but I can't ethically justify to myself doing anything else unless it doesn't change the outcomes.