Dear fellow researchers,
I have the following problem: I intend to conduct an experiment in an industrial environment where we expose the subjects to 2 treatments. We measure for each of these treatments several parameters (both subjective and objective, e.g. perceived workload).
In order to reduce the noise of the observations (e.g. a subject has a "bad day at work"), we intend to measure these parameters for each subject at multiple times, for instance twice a week. Multiple observations per individual provide a more confident understanding of one's perceptions and repercussing regarding his/her work environment.
I had the following questions:
i) Which statistical procedure should I use to assess the impact of the intervention? I was thinking about averaging the observations before the intervention per subject, as well as those after the intervention, such that I can perform a paired t-test (where each subjects measurement per treatment coonsists of its average score). However, in this way, I lose some information, and therefore I was thinking of a repeated measures ANOVA. Is the latter able to include both different treatments AND different time points? Or is another procedure even more suited for this particular analysis?
ii) Since the experiment is conducted in an industrial environment, no two tasks are exactly the same. However, the among-task differences are expected to be rather small and are not the scope of this research. This is partly addressed by observing several days, since this would level out the among-taks differences. Is this appropriate?
Best regards!
Nowadays, multilevel models (aka., mixed models or hierarchical linear models) provide a nice way to analyze repeated measures (or longitudinal) data. Try a Google search on , for example. And if you don't find something discussing your software, try again omitting ucla from the search term. HTH.
See the attached screenshot and download Prof M.Davidian course notes including code and examples. She developed most of this is which is standard in biostatistics Best wishes David Booth
To avoid the need for prey intervention and posts intervention data to be independent, a time series model may be formulated for the pre-intervention series. On the basis of the model forecasts are made for the posts intervention part. A transfer function is determined for the intervention, and the intervention model is obtained. Comparison is made between the post-intervention forecasts and data.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA