# [R] convergence for proportional odds model

From: liu abc <liu2074_at_yahoo.com>
Date: Tue 06 Sep 2005 - 00:42:39 EST

Hey, everyone,

I am using proportional odds model for ordinal responses in dose-response experiments. For some samll data, SAS can successfully provide estimators of the parameters, but the built-in function polr() in R fails. Would you like to tell me how to make some change so I can use polr() to obtain the estimators? Or anyone can give me a hint about the conditions for the existance of MLE in such a simple case? By the way, for the variable "resp" which must be ordered factor, how can I do it? Thanks a lot.

Guohui

The following is one example I used both in SAS and R.

in R:

library(MASS)
dose.resp = matrix( c(1,1,1,1,2,2,2,3,3,3, 2,2,3,3,4,4,5,4,5,5), ncol=2) colnames(dose.resp)= c("resp", "dose")
dose.resp
#> dose.resp
# resp dose
# [1,] 1 2
# [2,] 1 2
# [3,] 1 3
# [4,] 1 3
# [5,] 2 4
# [6,] 2 4
# [7,] 2 5
# [8,] 3 4
# [9,] 3 5
#[10,] 3 5
polr( factor(resp, ordered=T)~dose, data=dose.resp)
#Error in optim(start, fmin, gmin, method = "BFGS", hessian = Hess, ...) :
# initial value in 'vmmin' is not finite
#fitted probabilities numerically 0 or 1 occurred in:
#glm.fit(X, y1, wt, family = binomial(), offset = offset)

in SAS
NOTE: PROC LOGISTIC is fitting the cumulative logit model. The probabilities

```      modeled are summed over the responses having the lower Ordered Values in
the Response Profile table.
```

NOTE: Convergence criterion (GCONV=1E-8) satisfied. NOTE: There were 10 observations read from the data set WORK.ONE.