-1
$\begingroup$

I have 100 respondents from whom I have collected trip data for 3 months. I dont have control group or base condition.

I want to measure if the proportion of aggressive trip (Trips with atleast one aggressive event) vary between

  1. Weekday and weekend
  2. Morning and evening
  3. Low vs high speed trip
  4. Male and female
  5. Teen and adults

Can I use the dependent Paired t-test or Wilcoxon rank test (my data is nonnormal) for all?

$\endgroup$
  • $\begingroup$ One approach may be to arrange the data into a contingency table of counts: Yes-aggressive / No-agressive on one axis and e.g. Weekday / Weekend on the other axis. From there you could use a test like chi-square test of association. $\endgroup$ – Sal Mangiafico Dec 9 '19 at 16:48
  • $\begingroup$ If that makes sense for your data, you might update your question with you data arranged in this way. $\endgroup$ – Sal Mangiafico Dec 9 '19 at 16:54
  • $\begingroup$ You also might look at cross tabulation, which might provide some insight. $\endgroup$ – Sal Mangiafico Dec 9 '19 at 16:58
  • $\begingroup$ I think I cannot use Chi-sq test as the data is not independent. I have multiple trips from the same users. $\endgroup$ – MSilvy Dec 9 '19 at 17:07
  • $\begingroup$ It sounds like your dependent measurement is essentially dichotomous, so tests like t test or Wilcoxon won't make sense. One approach is to use logistic regression. If I understand your data, I don't think there's any simple test that would handle the structure of this data. $\endgroup$ – Sal Mangiafico Dec 9 '19 at 17:50
0
$\begingroup$

Actually the data is Naturalistic, so I don't have the same number of trips for all respondents. Thats why I am thinking of using the percentage of aggressive trips per driven mile or per trip as the dependent variable.

I think I can use Wilcoxon rank test for testing weekday/weekend or morning/evening etc and Chi-sq test for male/female test.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ As you say, depending on your data, it may make sense to calculate the percentage for each respondent, and then treat those as independent observations. I'm not sure why you are thinking about using Wilcoxon for one question and chi-square for another.... $\endgroup$ – Sal Mangiafico Dec 11 '19 at 19:25

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.