2
$\begingroup$

I'm performing a survival analysis and I have some quantitative variables that I categorized (but I also mantained the quantitative variable for the univariate analysis). For example, I have the variable age in terms of mean and SD and categorized (< 40yrs and >=40 years). My question is if I should include them both as covariates in the multivariable analysis or I should only include one of them (at the end, they're the same variable). If I had to include only one, which one would be better?

Thanks a lot in advance!

$\endgroup$

1 Answer 1

5
$\begingroup$

Categorization of continuous covariates is almost never a good idea. The reason is that you are basically throwing away valuable information, which may lead to a loss of statistical power or even biased estimates.

If your reason for categorization is that you want to graphically display your results, I would recommend specialized types of plots that directly use the continuous variable instead, such as survival area plots. See https://www.doi.org/10.1097/EDE.0000000000001630 for more information.

$\endgroup$
2
  • 1
    $\begingroup$ I can tell from the question that formal regression analysis and survival analysis study has not been undertaken. It is not wise to analyze data without the needed background knowledge. $\endgroup$ Commented Feb 11 at 14:25
  • 2
    $\begingroup$ +100 if I could for @FrankHarrell 's statement. But try convincing people! You wouldn't have surgery with someone who had one course in anatomy, so why do people do data analysis after having one course in statistics? $\endgroup$
    – Peter Flom
    Commented Feb 11 at 14:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.