I am trying to reproduce Hosmer & Lemeshow's example (3rd edition, Table 4.16, p115). With 'logistic' I get the same coefficients, Std. Err., etc. that they do, but if I then run 'ldev' I get number of covariate patterns = 471 whereas the text reports J=457 (p185) and so does 'estat gof'; also, I get a deviance from ldev of 474.50 whereas the text shows D=469.6 (p185) (the D value given by fitstat on the same sample is 500.5).
I use the ldev from ucla (1999); the other (sqv4 & sqv4_1 from stb) produces an error message ("factor-variable and time-series operators not allowed"), as does 'lpredict dev'.
Any help would be greatly appreciated.