Hi all, I am doing a survey where I would like to see if there is any difference in the amount of money customers spent at place A, and place B, C, and D. As the survey will be distributed in place A, all respondents will answer the amount of money they have spent in place A. However, respondents can only answer the amount of money they have spent in place B, C, and D only if they have visited those places. Therefore, the places they have visited (independent variable) is NOT limited to place A to D as mutually exclusive categories, it is a multiple response where respondents could have visited multiple places.
Am I able to conduct a one-way ANOVA repeated measures test when there are 3 places involved by copying all responses where participants have visited three places and conduct the test? Or can I only conduct multiple paired sample t-tests for my purpose? (e.g. comparing money spent in place A vs B, place A vs C, place A vs D)
You might also consider a multilevel model - unlike the one-way ANOVA this doesn't require complete cells.
Hi Evan Toh. You seem to indicate your subjects did not all visit all places. Some may have visited (e.g.) A, B and D; others may only have visited A and C; some may have visited A, B, C, and D; etc.. If this is the case, then you indeed could run multiple paired t-tests ( plus, maybe, also repeated measures ANOVA test (e.g. when some subjects visited A, B and C)). You could correct for multiple tests, for example, with a Bonferroni correction, or some other less conservative method.
See, e.g.
Article Beyond Bonferroni: Less conservative analyses for conservati...
https://en.wikipedia.org/wiki/Multiple_comparisons_problem
Dear Evan Toh , that's a good question.
You should always (given some conditions) analyze the whole data together. You can still make the proper comparisons you want to under a repeated records analysis with all 4 places.
Given that you are interested in all pairwise comparisons with location A, you may use Dunnett's test within your ANOVA, since this multiple test procedure was designed for cases like yours: many-to-one comparisons.
By using all of the data together, you have a better estimate of the variance due to repeated records on the same individuals, and thus, more accurate results.
Also, by using this method, you are NOT required to have the same participants in different locations as you would have for a paired t-test. So, even if a participant only went to places A, B, and C, this participant can still be used in the data.
Hope this helped! Please let me know if you have any questions. Nick
You might also consider a multilevel model - unlike the one-way ANOVA this doesn't require complete cells.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA