Mantel-Haenszel Chi-Square Assumptions

[Glenys Lafrance]

Hi listers,

Perhaps someone could help with a clarification here.  My sample size is
250, but many of my subset dichotomies have very different subset sizes
(ie, 8, 240) so I encounter small cell counts (which is why I got the
Exact Test module.  My DV is ordinal with 3 categories.  I have
encountered the following assumptions but I wonder if they have evolved
with new versions of procedures/software.

a)both variables must be ordinal (M-H is not appropriate if one variable
is nominal)...does this imply that dummy coded dichotomous IVs should not
be used?

b)M-H chi-square should not be used with tables with small cell
counts...is this valid if we have SPSS exact tests installed?

There are posts on this listserv that suggest M-H can be used as an
alternate to Mann-Whitney when the ordinal variable has limited range (3-
4).  So, clarification would help me alot.

Thanks in advance, I appreciate this group's collective wisdom.

Glenys Lafrance

Thanks for your consideration.

[Glenys Lafrance]

My apologies to Marta, who had already answered one of my queries.  The
Exact Tests module does permit analysis of tables with small expected cell
counts.  So I hope I can resolve the remaining question now.  Thanks.

Glenys Lafrance