Hi,
I’m doing a meta-analysis of the effect of two linear variables X and Z (with non-linear effect) on a survival outcome. I’m not having access to all data, so I cannot do IPD.
My model looks like this:
I would like to do a lincom test for the effect of various combinations of X and Z. My challenge is that the lincom must be run just after an estimation command (stcox), but since a meta-analysis can only handle one coefficient at a time, I have to do four meta-analyses (of x1, x2, y1 and y2).
My questions is now, if I can do the lincom test ‘manually’ after doing the four meta-analyses?
I tried to read the ADO-file for lincom, but I cannot find out what results I need to save from the stcox or from the meta-analysis and how I may possibly calculate the lincom test.
I can figure out to calculate the coefficient from combinations of x1, x2, y1 and y2, but does anybody know how I can calculate the std. err and the z-value with p-value? If possible at all? Does the metan/admetan command even store the numbers I would need for the lincom test?
Thank you
Lise
I’m doing a meta-analysis of the effect of two linear variables X and Z (with non-linear effect) on a survival outcome. I’m not having access to all data, so I cannot do IPD.
My model looks like this:
Code:
mkspline x1 15 x2 =y mkspline y1 30 y2=y stcox x1 x2 y1 y2, str(yy) nohr
My questions is now, if I can do the lincom test ‘manually’ after doing the four meta-analyses?
I tried to read the ADO-file for lincom, but I cannot find out what results I need to save from the stcox or from the meta-analysis and how I may possibly calculate the lincom test.
I can figure out to calculate the coefficient from combinations of x1, x2, y1 and y2, but does anybody know how I can calculate the std. err and the z-value with p-value? If possible at all? Does the metan/admetan command even store the numbers I would need for the lincom test?
Thank you
Lise
