hi
I am using spss and stata
I have experienced a problem with weighted analysis - the weights have been
set in spss and stata
I run either a crosstab in spss and include the line in the syntax (spss)
count=asis ,or svytab in stata, this will produce identical tables and
identical column percentages
The problem that I have is that the test statistic are different, I can
obtain a Pearson chi-square in spss, and also a p value based on the log
likelihood, but when I ask stata to produce the same I get a slightly
different value and different p value
Can anyone explain what is happening and if I should be using the log
likelihood values instead of Pearson in spss
The syntax for spps is below- I have previously set the weighting variable
CROSSTABS
/TABLES=wltind5a BY sib_grp
/FORMAT= aVALUE TABLES
/STATISTIC=all
/Count = asis
/CELLS= COUNT COLUMN.
The command for stata after using svyset to set the weights
svytab wltind5a sib_grp if filter__==1,count col pearson
The results from spss using Pearson give a pvalue of 0.032 and the log
likelihood p value of 0.044
The results from stata (which bases the result on a f distribution) gives a
pvalue of 0.045
