Re: st: Interpreting effect of covariate with min 0.2 and max 1.04 on odds/time ratios in PH and AFT models

[Maarten Buis <maartenlbuis@gmail.com>]

On Mon, Aug 15, 2011 at 10:40 AM, Erik Aadland wrote:
> I am estimating Weibull PH and AFT models.
> One of my timevarying covariates has a min of 0.2 and a max of 1.04.
> I want to interpret the effect of different values on this variable on the odds/time ratio.
> The values on this covariate are 0.2, 0.4, 0.6, 0.8, 0.9 and 1.04.
>
> The odds/time ratio reported is for each one unit increase in this covariate (I take it the one unit increase is 1).
> A one unit increase (1) does not reflect the values on my covariate well.
> Is there a way to get around this problem?

I am assuming you want hazard ratios rather than odds ratios.

One solution is to create a new variable which is the old variable
multiplied by 5, and add that to your model.

However, you do not need to evaluate the hazard ratio or time ratio at
different values, they will, by assumption, be the same whichever
value you choose, a unit increase in x will always lead to an increase
in the expected time/hazard by a ratio of exp(b). This is similar to
the linearity assumption in linear regression: a unit increase in x
will always lead b units increase in the expected value of y. These
assumptions do not have to be true, but you are not going to find
deviations from that assumption by inspecting a model that enforces
that assumption.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/