Hi everyone and thanks in advance for your help.
I am conducting a logistic multilevel model on a long-format dataset in which I want to evaluate whether the appropriateness of a drug prescription (binary outcome appropriateness) depends on the type of pathology (5 pathologies considered).
More precisely: I have about 600 doctors and I ask each of them if he prescribes the drug in the presence of each of the 5 pathologies considered (every doctor gives 5 answers).
In the model (melogit) I corrected for 5 dichotomous variables (age group, sex, type of training, type of experience, knowledge of prescription guidelines).
Obviously these 5 confounders are correlated very often (I made chi squares to verify).
I wonder if:
-I have to do the chi-squared tests on the long or wide dataset to verify the multicollinearity;
-if multicollinearity is a real problem in this case;
-does it make sense to test the chi-square for the cross factors, that is, the 5 pathologies considered?
Let me explain: the distribution of predictors (age group, sex, type of training, type of experience, knowledge of prescription guidelines) is inevitably identical for each level (i.e. for each of the 5 pathologies).
Finally, I wonder what techniques can be used in the STATA to correct the analysis and limit the effect of multicollinearity on the estimates.
Thank you very much!
Gianfranco
I am conducting a logistic multilevel model on a long-format dataset in which I want to evaluate whether the appropriateness of a drug prescription (binary outcome appropriateness) depends on the type of pathology (5 pathologies considered).
More precisely: I have about 600 doctors and I ask each of them if he prescribes the drug in the presence of each of the 5 pathologies considered (every doctor gives 5 answers).
In the model (melogit) I corrected for 5 dichotomous variables (age group, sex, type of training, type of experience, knowledge of prescription guidelines).
Obviously these 5 confounders are correlated very often (I made chi squares to verify).
I wonder if:
-I have to do the chi-squared tests on the long or wide dataset to verify the multicollinearity;
-if multicollinearity is a real problem in this case;
-does it make sense to test the chi-square for the cross factors, that is, the 5 pathologies considered?
Let me explain: the distribution of predictors (age group, sex, type of training, type of experience, knowledge of prescription guidelines) is inevitably identical for each level (i.e. for each of the 5 pathologies).
Finally, I wonder what techniques can be used in the STATA to correct the analysis and limit the effect of multicollinearity on the estimates.
Thank you very much!
Gianfranco
