## Contigency tables

Discuss statistics related things
fibankuti
Posts: 3
Joined: Wed Apr 10, 2024 6:21 pm

### Contigency tables

Dear all,
I will be pleased if someone can help-me. I would like to use contigency tables appling comparative mesures qui-square (3x4 ). But in Jamovi there is only compartive 2x2.
Someone knows how to do it? is there some module for that?

Thank you!
MattC
Posts: 8
Joined: Tue Jan 23, 2024 3:21 pm

### Re: Contigency tables

You could try using Logistic regression, which will produce odds-ratios and other related results. For a 3x4, you would need to select N-outcomes (Multinomial), and put one variable (probably the 3-level) as the dependent and the other as a factor. Interpretation of the results will be tricky though and needs careful thought to understand what each figure represents.
fibankuti
Posts: 3
Joined: Wed Apr 10, 2024 6:21 pm

### Re: Contigency tables

Thanks so much MattC.
reason180
Posts: 286
Joined: Mon Jul 24, 2017 4:56 pm

### Re: Contigency tables

Hi. jamovi's chi square contingency analyses aren't limited to 2 by 2. 3 by 4 works just fine.
three by four, temp.png (100 KiB) Viewed 1137 times
MattC
Posts: 8
Joined: Tue Jan 23, 2024 3:21 pm

### Re: Contigency tables

reason180 wrote: Fri Apr 12, 2024 2:55 am Hi. jamovi's chi square contingency analyses aren't limited to 2 by 2. 3 by 4 works just fine.
That was my first thought, but the original query was specifically about the "comparative measures" output (odds ratios, etc.) that are only available for 2x2 tables. I couldn't see any other way to do this in jamovi.
reason180
Posts: 286
Joined: Mon Jul 24, 2017 4:56 pm

### Re: Contigency tables

Oh. Sorry. However, I think those comparative measures are only computable in situation where the analysis can be conceptualized as comparing two things (to proportions, two ratios, etc.). Thus, they only work for 2 by 2 tables:

https://cran.r-project.org/web/packages ... square.pdf

I've seen computational procedures that automatically break a greater-than-two-by-two table into two-by-twos for purposes of calculating comparatives, but one could also just perform multiple analyses: the original n by n, along with the two-by-twos.
fibankuti
Posts: 3
Joined: Wed Apr 10, 2024 6:21 pm

### Re: Contigency tables

Dear reason180, thank you so much. I really apreciatte it.