Dear Allstatters
If I have a 2 by 2 contingency table with paired data, obviously I can
assess significance by McNemars test. In this case I am looking at presence
or absence of a sympton at baseline and after a set amount of time. If I
then wanted to assess two treatment effects on these symptons, i.e. I have
2, 2 by 2 tables with
this information, can I then assess the difference between the two without
resorting to a generalised model? This can obviously be done with unpaired
data through Mantel-Haentzel and odds ratios, but I have heard there is a
method to use with paired data.
Can anyone help?
Many thanks
Julie Sadler
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|