In an anatomic study, we suspect that our length measurement of an entity (1) is confounded by the length measurement of another entity (2). Entity 2, unlike variables such as age, smoking pack-year, etc. , has yet to be "stratified" into categories in the literature.
Before we attempt to make arbitrary categorization of entity 2, is there a way to disentangle the influence of entity 2 in the measurement of entity 1? Can this be performed through multiple linear regression?
if entity 2 is a continuous variable (eg length in cm) you could treat it as a covariate in multiple linear regression or in analysis of covariance .
I agree with James Leigh . Additionally, you may run your full multiple linear regression model. By plotting the residuals from this model against entity (2), you may asses if 1) it is appropiate to treat it as a continuos variable (with linear realtionship to the outcome variable) and 2) which categories it is appropiate to devide the variable into( or how to "stratified" it, in your terminology). I can further reccomend to add a lowess smoothed curve to the plot. A more simple approach could be to simply plot the outcome against entity 2 and add a lowess smothed curve this might suggest relevant cut-points.
Yes, you can use the multiple regression. Don't forget to make a scatter plot to show whether the correlation is linear or not.
- Badges
- Science topic
- Similar topics
- Philosophy
- Ethics