| |
 |
|
|
Science Forum Index » Statistics - Education Forum » Cox Proportionmal Hazard Model
Page 1 of 1
|
| Author |
Message |
| Guest |
 Posted: Tue Jan 02, 2007 1:14 pm |
|
|
|
|
I am more than a little confused by the figures produced by the Cox
Proportioal Hazard Model (specifically when using SPSS) and would like
some advice (or possibly insults as to my stupidity)
I have a groups of subjects who have self-rated themseleves as
programmers on a scale of 1 to 10 and then asked to complete a
programming task. I measure how longit takes them to compete the task
and they get a maximum of 60 minutes.
So this is stright forward, it's right censored data, with a continuous
independent variable. I plug my numbers inot SPSS and it says the
p-value is 0.021. So that means it is an indicator of completion time
yes? The point where I get confused is I also have their grade for
their programming class, if I use that as a variable then it too is a
significant indicator. However, and this is where I do the Cox
analysis for both of them together then it says that the self-rating
_isn't_ a significant indicator but the grade is.
Could someone explain (or point me to the explination) of what's going
on and what that means I can say about the subject's self-rating? |
|
|
| Back to top |
|
| J W |
| Posted: Wed Jan 03, 2007 2:45 am |
|
|
|
Guest
|
Quote:
So this is stright forward, it's right censored data, with a continuous
independent variable. I plug my numbers inot SPSS and it says the
p-value is 0.021. So that means it is an indicator of completion time
yes? The point where I get confused is I also have their grade for
their programming class, if I use that as a variable then it too is a
significant indicator. However, and this is where I do the Cox
analysis for both of them together then it says that the self-rating
_isn't_ a significant indicator but the grade is.
Could someone explain (or point me to the explination) of what's going
on and what that means I can say about the subject's self-rating?
You could say that after controlling for self-rating, grade remains a
significant predictor of completion time. Alternatively, you could say
that after controlling for class grade, self-rating is not a significant
predictor of completion time. Which phrasing you choose should depend
on which of the two predictor variables is of primary interest (from
what you indicate, it sounds like self-rating is of primary interest,
and hence the second sentence may be more appropriate).
As for "what's going on", it's hard to say with any certainty without
access to the data and further analyses. It could be the case that
grade is simply a much more important predictor of completion time than
self-rating, and hence the effect of self-rating is masked by putting
this other variable in the model. Also, if grade is strongly associated
with self-rating, then the univariate model including self-rating as the
only predictor may be measuring some combination of the effect of grade
and self-rating; the bivariate model will "separate" these effects.
Hope this helps.
- J |
|
|
| Back to top |
|
| Guest |
| Posted: Wed Jan 03, 2007 4:01 pm |
|
|
|
|
J W wrote:
Quote:
As for "what's going on", it's hard to say with any certainty without
access to the data and further analyses. It could be the case that
grade is simply a much more important predictor of completion time than
self-rating, and hence the effect of self-rating is masked by putting
this other variable in the model. Also, if grade is strongly associated
with self-rating, then the univariate model including self-rating as the
only predictor may be measuring some combination of the effect of grade
and self-rating; the bivariate model will "separate" these effects.
Thanks for your help, I'd say that's pretty much what's going on, each
grade band (A,B & C) had it's own rough range of self-ratings but
there's not much difference between the means of each group (7 for the
As, 6.5 for the Bs and 5.7 for the Cs) becasue people tended towards
rating themselves as a 7.
--
Alistair Hutton |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Tue Oct 07, 2008 11:48 pm
|
|