I am a bit confused about this question. I am currently attempting to do a meta-analysis on pesticide exposure and telomere length (outcome). The exposure(s) were basically presented as either a continuous level of pesticide in a biological sample or a dichotomy(exposure/no exposure) using a questionnaire. Both of the estimates were beta coefficients. My question is: can I combine them to do a meta-analysis? If so, how to interpret the results?
Thanks.
I think you'll have to perform 2 different models (or select what is the more realistic/pertinent approach). The beta obtained from your models tells you different things: for continuous measure of exposure this gives you the impact of each unit increase of exposure and the dependent variable (telomere length).
For pesticide as a dichotomous outcome the beta gives you the effect of presence of exposure (independently of its importance) on telomere length.
It makes no sense to put the different studies in the same model because of these differences (like comparing pear and apple).
You could also convert exposure level to a dichotomous variable (perhaps using some threshold value) - you loose some information, but it allows combining the sample sizes. Then add an analysis of the subset with continuous data available and compare the results.
Remember that there are a lot of other biological and environmental factors that may be more relevant to shaping telomere length than pesticide exposure.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis