survey - how to deal with replicate weights equal zero in R? -
i have 0 values replicate , sampling weights. therefore, when use svycoxph “survey” package, error message “package invalid weights, must >0”. think 1 way might exclude these observations. wonder if there way keep observations cox proportional hazards model?
thanks! julia
zero replicate weights normal -- jackknife replicates observations psu being left out; bootstrap replicates psus appeared 0 times in particular resample. 0 sampling weights, on other hand, don't make lot of sense , indicate observations aren't in sample or aren't in sampling frame (personally, i'd code these na
)
the coxph()
function in survival
package can't handle negative or 0 weights, , that's svycoxph()
calls. replicate weights, svycoxph()
adds small value (1e-10) weights stop problems zeros.
however, svycoxph()
can't handle 0 sampling weights. want remove these before constructing design object.
Comments
Post a Comment