Hello everyone,
I have discovered that for logistic regression (binary outcome situation,
*ungrouped* case i.e. ni=1), it is not valid to use the scaled deviance or
Pearson's chi-square to assess goodness-of-fit. The "Hosmer-Lemeshow" test
is instead recommended.
For ordinal logistic regression, is there a similar test to assess goodness
of fit in the 'ungrouped' situation? If so, is there a command in
SPSS/Minitab/SAS which is suitable for the task?
many thanks,
Kim
[log in to unmask]
|