Cox regression problem
Locked
 
|
I need to compare the hazard ratios of two treatments A and B within two subgroups defined by risk (high or low). My variables are
- Time (to event or censoring), - Event (whether or not the event of interest has occurred), - Treatment (A or B), and - Risk (high or low). The following information is required: - The hazard ratio for treatment A vs. B. within each subgroup defined by risk (point estimate, confidence interval, p-value, and - whether the treatment hazard ratios differ between the high risk and the low risk subgroup (point estimate and confidence interval for difference, p-value). How can I specify this analysis in Cox regression? Thanks very much! Andreas |
|
|
So, two groups differing in risk level (LR and HR) and in each group some people receive treatment A and the rest of the people receive treatment B. Why isn't this just an interaction problem: A=0 vs B=1, LR=0 vs HR=1 plus their interaction?
Gene Maguin -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Andreas Voelp Sent: Monday, July 06, 2015 8:00 AM To: [hidden email] Subject: Cox regression problem I need to compare the hazard ratios of two treatments A and B within two subgroups defined by risk (high or low). My variables are - Time (to event or censoring), - Event (whether or not the event of interest has occurred), - Treatment (A or B), and - Risk (high or low). The following information is required: - The hazard ratio for treatment A vs. B. within each subgroup defined by risk (point estimate, confidence interval, p-value, and - whether the treatment hazard ratios differ between the high risk and the low risk subgroup (point estimate and confidence interval for difference, p-value). How can I specify this analysis in Cox regression? Thanks very much! Andreas -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Cox-regression-problem-tp5730066.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
... yes, I agree that the test whether the treatment hazard ratios differ between the high risk and the low risk subgroup is of course in the interaction between the two predictors. However, how can I display the hazard ratio for treatment A vs. B. within each subgroup defined by risk (one would have to hold risk constant at one level and then request the hazard ratio for treatment)? I can think of a workaround (use the Split File option to perform the regression analysis within the risk subsets), but this is apparently not very 'elegant', and I assume that there must be a way of specifying the required contrasts within just one analysis.
Andreas |
|
|
Sounds like you'd like to have planned comparisons as in GLM/Mixed. Those aren't available. Rather than the inelegance of split files how about the Strata subcommand? Gene Maguin
-----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Andreas Voelp Sent: Tuesday, July 07, 2015 3:42 AM To: [hidden email] Subject: Re: Cox regression problem ... yes, I agree that the test whether the treatment hazard ratios differ between the high risk and the low risk subgroup is of course in the interaction between the two predictors. However, how can I display the hazard ratio for treatment A vs. B. within each subgroup defined by risk (one would have to hold risk constant at one level and then request the hazard ratio for treatment)? I can think of a workaround (use the Split File option to perform the regression analysis within the risk subsets), but this is apparently not very 'elegant', and I assume that there must be a way of specifying the required contrasts within just one analysis. Andreas -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Cox-regression-problem-tp5730066p5730078.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
|
In reply to this post by Frank Furter
Andreas, After quickly perusing SPSS documentation, it appears as though the CONTRAST sub-command along with the SPECIAL keyword would allow you to generate the contrast of interest. Ryan On Mon, Jul 6, 2015 at 7:59 AM, Andreas Voelp <[hidden email]> wrote: I need to compare the hazard ratios of two treatments A and B within two |
|
|
Haven't found a way to display the hazard ratios for my two sunsets with either high and low risk in just one analysis. So it appears that my workaround using the Split File option for p0erforming separate Cox regression analyses for each of the subsets wasn't so bad after all. Then I need to add another Cox regression with Treatment and Risk as predictors as well as the interaction between the two. The latter will tell me to which degree the effect of Treatment is modified by Risk, right?
Thanks, Andreas |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |