Hi I thought I'd use this forum again
I performed a Cox analysis on: If allied health professional performing specific operations using an orthopaedic implant have better outcomes if they do more of these interventions
I then wanted to assess PH - so I got a GOF using phtest, plotted scaled schoefield.
The only variable that didn't satisfy PH was gender (graph of schoefield residuals seen in this post)
https://www.statalist.org/forums/for...esiduals-graph
I then introduced a time varying covariate option
Measured gof, scaled schoefeld - and PH were satisfied..... hurrah!
However, rather than just performing Statistical tests I like to understand what's going on. Here are a couple of questions
My question:
1.How would one interpret introducing a time varying option for PH in a orthopaedic implant model
Is it the risk of failure for each gender varies over time ? - I've just been through 77 pages of PUB med results to try explain this
2. I know it's possible to plot this graph with R, is it possible to do so with STATA?:
2. According to Modelling Survival Data by Terry tHERNEAU - Chapter 6.5
He gives two option to either stratify or introduce a TVC.
I'm not sure which option is better.
I know that by stratification, the model is generated from the non-stratified variables and the variable gender is not used as a regressor.
Can anyone expand on this? -
Just trying to full understand what I'm doing.
I'm sure there are people on here who can explain this better, hopefully there are some stats people around...
I performed a Cox analysis on: If allied health professional performing specific operations using an orthopaedic implant have better outcomes if they do more of these interventions
Code:
stcox var1 var2 var3 gender, nohr
The only variable that didn't satisfy PH was gender (graph of schoefield residuals seen in this post)
https://www.statalist.org/forums/for...esiduals-graph
I then introduced a time varying covariate option
Code:
stcox var1 var2 var3, tvc(gender) nohr estimate storem1 stcox var1 var2 var3 gender, nohr //original lrtest (storem1) //significant
However, rather than just performing Statistical tests I like to understand what's going on. Here are a couple of questions
My question:
1.How would one interpret introducing a time varying option for PH in a orthopaedic implant model
Is it the risk of failure for each gender varies over time ? - I've just been through 77 pages of PUB med results to try explain this
2. I know it's possible to plot this graph with R, is it possible to do so with STATA?:
2. According to Modelling Survival Data by Terry tHERNEAU - Chapter 6.5
He gives two option to either stratify or introduce a TVC.
I'm not sure which option is better.
I know that by stratification, the model is generated from the non-stratified variables and the variable gender is not used as a regressor.
Can anyone expand on this? -
Just trying to full understand what I'm doing.
I'm sure there are people on here who can explain this better, hopefully there are some stats people around...
Comment