i have this problem to solve. I have a questionnaire of 19 questions, each one has 3 options, and only one for each is correct and it scores 1, while the others scores 0. Thus the max score will be 19. I have given the test to 100 persons and in addition i have collected some demographic data such as gender (M or F), age (<30, 30-40, >40), experience in the field (<5y, 5-10 y, >10 y), experience with a particular issue (Y or N) and others. Now i want to see if any of the demographic factors could influence the score. What tool should i use? I thought of a categorical regression (in SPSS, even though i'm not in SPSS but my client want me to use this), since some of the demographic data are categorical, such as gender, experience etc. If so, i have run some regressions using the CATREG options under Analyze in SPSS, but i'm not sure how to define the scale for each variable. Or maybe i should go for ordinal regression if the variable is ordinal such as age or experience in the field, and categorical regression otherwise (for variable such as gender or experience on a particular issue. Any comments would be appreciated. Thanks in advance
The score that respondents receive is a sum of 19 0/1 outcomes. With those numbers, the total score is probably a decent approximation to a normal distribution, so I would just go for multiple regression. SPSS can handle categorical independent variables.
If you want to get fancy with the categorical IV's, and make use of their ordinal character, you can use the contrast option in SPSS and request polynomial contrasts. For example, with a category that has 3 levels, like age, you can supply contrast (-1,0,1) to estimate a monotone effect and (1,0,1) to estimate departures therefrom.