Variable 1 (education levels)
Variable 2 (yes/no)
I want to check if there is an association between education level and ability to complete a task (yes/no). I would normally go for a simple chi squared test, however I want to repeat the experiment using the same participants but with five variations on the tasks, checking for an association.
Is it ok to perform 5 seperate chi squared tests to check for an association between education level and each of the five tasks ?
Or is it better to total the score from the five tasks and check for an association between education level and the combined tasks.
This is an interesting question. You could use the two variables as mentioned (education and task completion) and create a different categorical variable (say task variety, with five different categories), then run a log-linear analysis to see if a two (between education and task completion) or three-way association exists. If only two way, then the type of task does not influence the association between education and task completion. If there is a three-way, then you can follow up with separate chi-squares to see where the differences lie. Does that make sense?
Andy Field has a good section in his book describing Log-Linear analysis.
Lee John Curley thanks for the advice Lee. For the new variable do you mean for each of the five tasks (Yes/No) whenever someone scores "Yes" assign that a category on the new variable ?
Hi Christopher, what I mean is you could have three variables: variable 1 (education level); variable 2 (completion of a task; yes vs. no); and variable three (task 1; task 2; task 3; task 4; and task 5). Then by using the Log-linear analysis mentioned above, you could see if the association between education level and completion of the task is influenced by variable 3. Does that make more sense?
Lee John Curley ah I see now that makes sense. Ok great im going to look into that. Thank you for your help
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts