Solution for Non-normally distributed data

[Agate]

hello there,
I am doing some statistical analysis on medical sample data.
I need to need to find variables which influences the length of stay in hospital.
I was panning to do collaboration analysis and stepwise model for detecting regression model.

My data is nor normally distributed? How critical is that?

And how can I normalize it??

Thanks
Best regards
Agate

 

[trinker]

My data is nor normally distributed? How critical is that?

The assumption is that the error terms are normally distributed (i.e., the population is normally distributed). Look at a QQ plot of the residuals.

 

[Karabiner]

Or, length-of-stay data may perhaps be better modeled
using survival analysis (e.g. Kaplan-Meier, or Cox regression).
Unfortunately, one of the most crucial information (sample
size) wasn't mentioned, but please stay away from stepwise
at any case http://www.stata.com/support/faqs/statistics/stepwise-regression-problems/

With kind regards

K.