Given the repeated nature of the study examining different cohorts each time, how can one make comparisons of eating specific food with implications toward changes in eating behavior?
"repeated cross-sectional” study in which the same survey is conducted over time and is directed toward a new sample of different participants at consecutive time points with regular intervals.
Would it not be possible to do a semi experimental study.
eg take two groups of volunteers and give each a fixed food regime for a fixed period .( One group would be a "control" group with a regime you would not expect to alter future eating patterns.) .
Then compare future eating patterns in each group over a further observation period.
I take it that you are analysing data already collected.
To start with I would stack each cross section underneath each other , responses and predictors and a code for each time period. I would then create a dummy for each time period and then fit a regressio—like model with no predictors( a null model) then add in the dummies (has the response changed over time?) then add relevant predictors ( control and regime) and appropriate co factors (age sex, .....) to the see the differences. And finally time dummy by predictor interaction to see effect has change over time. What you can do in this exploratory manner depends on power and theoretical nous. If you have a big dataset consider a holdout sample from the outset. One to explore , one to confirm. Good luck.
i know nothing about the practicalities about measurement but clearly It needs to be consistent over time.
kelvyn
My understanding from my data sets is that the main drawback of repeated cross sectional study is that the participants are different in each time of surveys and not followed over time Thus, through the repeated cross-sectional survey, it is not possible to evaluate changes over time at the individual level. But it makes it possible to compare the population and to analyze changes over time in population or groups as aggregate change.
Yes that is the case and given current knowledge that may still may be an important contribution especially if you test against standard theory .
Thank you prof.Jones for your comments. I also think that aggregate data for a geographic unit as a whole may be in this case only weakly related to the diets of those individuals at risk of specific disease.
I have access to the diets individual data but not their health outcomes.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey