0
$\begingroup$

In our research on apprenticeships in early modern times, we have a series of observations (cases), each with several variables. Each observation is unique because of its combination of the variables "student name" and "starting date of the apprenticeship" (a certain "student name" can occur in more than one observation, a certain "starting date of apprenticeship" may also occur in more than one observation, but a certain combination of the two can only occur in one observation - compare with primary key in Access).

We want to investigate the relationship between various other variables ("status apprentice", "parents deceased", "guild registration", "duration apprenticeship", "wages", "who pays the wages", "apprenticeship completed") by means of SPSS Crosstabs + chi-square. Variables can be dichotomous, nominal, ordinal or scale (binned to nominal categories). Later on we intend to perform Logistic Regression Analysis.

The fact that different cases can have the same apprentice might suggest that these cases are paired and that chi-square might not be appropriate. On the other hand, the values ​​of other variables can also occur more than once, and you could say that the observations are also linked by those variables. Or by extension: in a large dataset the values ​​of one or more variables often occur more than once, so that you will rarely see a dataset with only unpaired observations?

Can we perform Crosstabs + chi-square on this data? Can we perform Logistic Regression Analysis on this data?

$\endgroup$
0
$\begingroup$

If you want to investigate the effects of several variables (in your case "status apprentice", "parents deceased", "guild registration") on a categorical variable, then logistic (or ordinal) regression seems to be appropriate. However you will first need to clarify your research question - What would be the dependent variable of the model? If you only want to investigate the relationship between 2 variables, then logistic regression would be over-killed and simple chi-2 tests would do the job. If the data are not independent then you could use McNemar test as an alternative to chi-2.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.