This error appears on Windows since the release of optimx version 2018-7.10. The issue was raised on Stackoverflow, see https://stackoverflow.com/questions/52927399/renewalcount-function-in-countr.

A workaround for users is to do xtimes <- 1 before calling renewalCount(). Another, more drastic, option is to go back to the previous release of optmix.

I am opening the issue here for easy reference. The example below is from the Stackoverflow link above (I prepended an option() command for debugging).

options(error = recover)
library("Countr")
library("dplyr")
library("xtable")
D=matrix(c(1:9))
Numb=matrix(c(15, 4, 5, 3, 1, 2, 1, 1, 1))
Number=data.frame(Numb)
a_weib <- renewalCount(formula = D ~ 1, data = Number, dist = "weibull", 
     computeHessian = FALSE,  control = renewal.control(trace = 0))

Running the above code (under Windows) gives:

Error in order(-value, xtimes) : object 'xtimes' not found

xtimes is documented to be part of the result of 'optimx' and fitCount_ is the object returned by optimx(). Printing it gives:

Browse[2]> fitCount_
         scale_     shape_     value fevals gevals niter convcode kkt1 kkt2
nlminb 1.343784 -0.2194413 -21.21925     12     21     9        0   NA   NA
       xtime
nlminb  0.34

We see that there is a column xtime, not xtimes. Indeed,

Browse[3]> names(fitCount_)
 [1] "scale_"   "shape_"   "value"    "fevals"   "gevals"   "niter"   
 [7] "convcode" "kkt1"     "kkt2"     "xtime"   

The problem may be ocuring in optimx:::optimx.run(). It assigns ans$xtimes <- time on line 736 of optimx.run.R but later assembles another object from ans and other stuff, without explicitly mentioning column names.

The line of code in Countr where the error occurs needs to be replaced in any case by a more robust version.

I have a database and I would like to use the gamma and weibull distribution to model the number of events that occurred during different time intervals.

I have found the countr package to meet my requirements perfectly. I made two scenarios, in the first one I model the number of events according to some covariates without using the offset parameter. In the second scenario I used the offset parameter. Comparing summary of the two scenarios, the result is identical. This result does not match my expectations and I do not know why?

Regarding the Countr package, I would like to know if the offset parameter is implemented in the renewalCount function ?

I am a new user of rstudio.

Thanks

I don't have any NA in my data but I have this error and I don't understand the meaning

Error in if (var < 0) { : missing value where TRUE / FALSE needed
De plus : Warning message:
In .checkInitialValues(dist, start, modelMatrixList, weights, Y, :
unnamed initial values found! They will be used as they are.

Thanks