# A non-parametric ANOVA for nominal data

#### capri87

##### New Member
Hi,
I have a research project that has two groups, I'm trying to analyse people's response to an open ended question that I have coded according to five themes. So, when I input the data into SPSS, each row has what group they are in and the theme (1-5) that they have. Would anyone know how to analyse this data using an ANOVA? I've been going over my stats books over and over again but i just can't understand how to do it! :shakehead
Thank you so much for helping!

#### Lazar

##### Phineas Packard
A few quick questions:
1. Are the open ended questions rated on the number of times each theme emerges (e.g. they got a count of 2 for theme 1 but a count of 3 for theme 2 etc.) OR was the response coded such that each individual had a single code (1,2,3,4,5) that indicated which them they responded to.
2. What is your reason for wanting to do ANOVA?

If I have understood you correctly you have groups as the IV and theme presence as the DV (where each individual gets a single code of 1,2,3,4, or 5=. As such is your data not set up for a cross tab something like this:

Group theme1 theme2 theme3 theme4 theme5
1..........6...........9..........2...........4.........5
2..........7...........1.........10.........10.........2

#### capri87

##### New Member
Hi,
Thank you for replying.
In response to your questions: 1) Each individual has a single code (1-5) and 2) My supervisor has suggested that I should conduct an ANOVA.
I'm using SPSS to analyse my data, and I enter the data like so:

Group Qn X Qn Y
1 2 1
2 5 1
1 1 3
1 4 2
2 3 4
2 2 1

I initially did a chi- square thinking that's most appropriate for this data, but I'm just trying to figure out how to do an ANOVA with this. Once again, thank you for replying.

#### Lazar

##### Phineas Packard
hmm not sure I understand 100%. Is QnX and QnY independent outcomes in your analysis that are both predicted by group. Is that the sort of thing you want to predict? If so chi-square is the approach I would have taken.

#### capri87

##### New Member
Hi,
My research is not an experimental one- Qn X and Qn Y are on their views on my research topic which is Financial Elder Abuse. So, I've got 2 groups based on their culture and Qn X and Y are open ended questions. I've coded the responses to each question and I'm trying to analyse if the responses to the questions are different between the two groups. I'm so sorry if I'm confusing you, but thank you so much for your help.