Dear colleagues, In particular, my investigation is designed in the model of pre/post test single group statistical test (i think non parametric).Data was collected in two periods: before and after applying the intervention. The problems is this study was conducted on only one single group (single arm group) using the same set of questionnaire at two different periods. once before and one after the intervention.My question is , is the study design: intervention pre post study design ?
what will be statistical tests to check the difference between these two? (most questions options are on likert scale)
A t-test of the difference between the pre- and the post-intervention data means could be done to ascertain the difference. A more holistic approach is to fit an ARIMA model to the pre-intervention data as recommended by Box and Tiao(1975) who pioneered this approach. Then generate post-intervention forecasts on the basis of the model. Then model the difference between post-intervention observations and forecasts to get the intervention model. Significance of the model amounts to the significance of the pre- and post- intervention difference.
The appropriate statistical test will depend upon if you are able to pair the responses. That is, if you know how an individual answered on the pre-test and you can pair the pre-test response to the post-test for the same individual.
The design is experimental study or Intervention study. The test would be paired-t test. If you knew the test result (eg: total score) before and after intervention for each individual. If you don't know who got which score for which time (before or after) do a t-test .
I recommend against using t-test or anova for Likert item data. Likert item data does not meet the assumptions of these tests.
It's best to use ordinal regression. But I also think it's okay to use traditional nonparametric tests, tho some people disagree.
- Badges
- Science topic