obtain odds ratio from regression for continous variable

[Denise Vella]

Hi I would like to obtain odds ratio for a continuous dependent variable. All variables in the code are continuous variables
Charlson - is a continuous variable that measures the pt comorbidity

reg ptage splinevar* charlson [pw=weight]

This gives me a coefficient value
eg: charlson has a coef 0.06

My interpretation is: The sicker the patient the more likely the older the pt is by 0.06 points.

However can I get an odds ratio instead?
Does it make sense to get an odds ratio is perhaps I can interpret with the co efficient?

[Maarten Buis]

I assume that the variable ptage is a percentage. To get something like odds ratios you first need to transform the percentage to a proportion, i.e. divide by 100. After that you can use fracreg logit prop splinevar* charlson [pw=weight], or

[Denise Vella]

Apologies for not being clear
ptage is a continuous variable for ‘patient’s age’

so are you saying I should just

Logit var1 splinevar* var3 [pw=weight], or

[Jeff Wooldridge]

Why is patient's age the outcome variable? Or is it the age at death? With a continuous variable it doesn't make sense to discuss "odds ratios" -- at least, not to me.

[Denise Vella]

the splinevar* is a spline fitted on physician experience

i want to see if physicians with more experience treat patients who are older ?
in summary, with more physicians experience is there a difference in the type of patients the physician treat in terms of age

i have used spline to leave experience as a continuous variable rather than categorise it.

if you don’t agree with odds ratios in continuous variables, then my interpretation in post1 is correct in terms of corfficient

[Carlo Lazzaro]

Denise:
you can investigate if physicians' experience shows a turning point, that is a non-linear relationship with -ptage-:

regress ptage c.phys_exp##c.phys_exp <otherstuff>

In addition, I would investigate whether the default standard errors are appropriate and the functional form of the regressand is correctly specified (see -linktest-).
Eventually, to assess the correctness of your statement, sharring what you typed and what Stata gave you back (as per FAQ) can help enormously.