PAF through formula or through Conditional Logistic Regression

[Pure_Stats]

Hi all,

In a large registry dataset of psychiatric patients, if one wants to find the excess risk of schizophrenia patients developing depression, what would be the best method to apply:

1) Use the relevant formula to calculate PAF (Population Attributable Fraction).

Formula can be found here: http://www.who.int/healthinfo/global_burden_disease/metrics_paf/en/

OR 2) perform a conditional logistic regression (using the clogit command in stata) to obtain the excess odds of schizophrenia patients developing depression.

What is the key difference between the two (if any)?

Thanks in advance.

 

[Englund]

The PAF is a simple statistic and probably easier to interpret than the logit. But I would say that the logit model can accomplish a lot more than the PAF and I would go for that 9 times out of ten.

 

[Pure_Stats]

Hi, thanks for your response. But, what is the key difference between the two approaches?

Thanks