I'm running a few different analyses on data with a nominal IV's (eg exercise/ not, use a computer or not, previous neck injury or not) and interval DV (a measure of muscle strength ranging from 0-100 points) so I'd like to run a few t tests on my data. Is there a difference between the 2 types of t tests?
Hello Jeanine Blumenau. When I click on Analyze > General Linear Model > Univariate, the options I see are:
- Univariate
- Multivariate
- Repeated Measures
- Variance Components
I do not see a t-test option, so I am not sure what you mean when you say GLM-UNIVARIATE t-test analysis. You might mean that you are selecting the option to see a table of parameter estimates that includes t-tests on coefficients, but it is not clear. It would help clear things up if you shared your command syntax for both the ordinary t-test and the GLM-UNIVARIATE t-test you are talking about. If you have never used command syntax, you can generate the code by exiting the GUI dialog via the Paste button rather than OK. Thank you for clarifying.
One further comment. Judging by some of your dichotomous explanatory variables, I am going to guess that you have observational data, not experimental data with random assignment to groups. (E.g., you would not randomly assign people to have a previous neck injury or not!) Therefore, the group sizes are likely not the same, and may differ substantially. Bear in mind that the homogeneity of variance assumption becomes more important as the sample sizes diverge. Therefore, when the sample sizes are not equal (or close to equal), it is a good idea to use either the Welch or Satterthwaite t-test, as these do not require homogeneity of variance. When you use the T-TEST command in SPSS, the output includes a row for an "equal variances not assumed" test. It is the Satterthwaite t-test.
HTH.
Hello Jeanine Blumenau If your independent variables are binary like you said (eg exercise/ not, use a computer or not, previous neck injury or not), using either of the paths you said will give you the same results. Of course, GLM (Univariate) is not called t test. It is interesting to note that even if our independent variable is binary, it is not the case that we must use t-test but can still use Univariate. In particular, they are more useful and can also achieve Interaction-effects (if we look to check them, of course).
- Badges
- Science topic
- Similar topics
- Engineering
- Materials Engineering
- Powders