Crosstabs Interpretation?

#1
Hello, I'm trying to figure out how to interpret the proportion (Z test) calculated in the crosstab table in SPSS.

In the attachment, in crosstab 1, i'm trying to analyze graduates as my DV by module #. By the numbers it seems that there would be significant difference in proportion on Module 4 in row 1, and Module 1 in row 2. but it doesn't seem to be saying that?

So then I ran it again in crosstab 2, and switched the columns and rows, and then it looked right?

Could someone please explain why? THANK YOU!!!
 
#2
Or, does anyone know a better way to test the significance of the proportion/percent of a nominal DV on levels of a nominal IV? (my understanding is that chi-square measures the significance of actual cases?)
 
#4
Thank you for your response, I will look into the Correspondence Analysis now.

The data attached was data I made up to make sure I was interpreting the Z-test correctly-

My dissertation data has graduates (n=19) and dropouts (n=57) of a program, and I am examining the relationship between graduates and dropouts and the first module they started the program on; 1,2,3 or 4.

While the chi square was not significant, we felt that could be due to the unequal groups, and so that is why we wanted to look a the percent/proportion for significance as well.

Thank you!!!
 
#5
Ok, I've looked at CA, and I think, because my chi-square did not come out significant, this would not be helpful for what I'm trying to do.

But perhaps, I'm on a wild goose chase- is it possible that even though my chi square/cramer's V was not significant, that the percentages or proportions could be? Graphically, it appeard that a higher percentage of graduates attended Module 1. However nothing has come out significant, and I was led to believe that was because I was looking at individual cases.

I really appreciate help on this!! THANK YOU!!!
 

gianmarco

TS Contributor
#6
Hi!
Since I do not know in detail your dataset, I cannot be more precise in my answer. But, be adviced that CA (as exploratory approach) and statistical significance (i.e., chi-square) are two different stories. Relying upon what Greenacre says in his book, you can perform CA even if chi-sqaure is not significant. This as a general rule.

I think that, provided that your dataset is worth exploring, you could try to run CA anyway.

Hope this helps,
Regards
Gm