What fundamental concerns need to be kept in mind when we compute research analysis in the fields of Statistics, Biostatistics, and Epidemiology? Will we consider only significant variable for our model building? Will we think about the biological factors, whether it is substantial or not ? Is it necessary to check the assumptions for different models as some papers do not mention them? Were interaction terms always essential to check?
You must consider only significant variables with traditional statistics. You can, however, state there is a biological difference, even though results don't reach the statistical threshold. Assumptions of models should always be stated, especially if you are ascertaining models that best fit your results. And interactions must always be included, since type 1 errors will be introduced if you exclude them :)
The biggest actual problem is quality of data, often corrupted, often due to conflicts of interest.
Samuel Oluwaseun Adeyemo Thank you for your explanation. Moreover, Sometimes we need to think about the confounding effect and effect modifiers. Most of the paper articles do not consider them. Is there any impact of these in the final model?