I know many questions have been asked about this topic, and I have looked at the user manual.
First: my data. I have dichotomous data (yes/no) for a number of health related questions.
However, I still have trouble interpreting my tech10 output. I have read that a z-score of over 1.96 is significant (that makes sense), but I do not know which score to look at.
For every variable combination, there are 6 z-scores, 4 for all possible answer combinations (yes-no, yes-yes, no-no, no-yes), the bivariate person and the bivariate log-likelihood. Which one should I evaluate?
Also, I feel it should be relatively simple to allow for local dependence (say within a class that [smoking before age 13] is allowed to correlate with [smoking for 30 days in a row] in a class.
I have added "algorithm = integration;" to the ANALYSIS statement, and the model part: Model: %OVERALL% [sm13] by [sm30days];
It tells me "Covariances for categorical ... are not defined". I do not understand why I get this message.