I need to analyse some data in IBM SPSS statistics 21. I'm a novice to statistical testing and hope someone could enlighten me.
Research is looking into the relationship between breed size and how an owner approaches behavioural problems.
A closed question survey was used to collect data from 500 owners. Purebred dogs were only used for size categories (small, medium, large).
There were unequal total responses:
Small: 201
Medium: 133
Large:166
All my questions are typically similar to this in layout (main question, followed by a “how do you manage this?”).
Example question data from survey:
H0: There is no relationship between breed size and owners' approach to aggression in play.
H1: There is a relationship between breed size and owners' approach to aggression in play.
6a. Does your dog have play aggression?
Small: Yes = 32. No = 169.
Medium: Yes = 17. No = 116.
Large: Yes = 16. No = 150.
6b. If yes, how does your dog manage this?
1. I don’t have anything in place to manage this. I don’t think it’s a problem.
Small:12
Medium: 5
Small:12
Large:1
2. I stop play when my dog gets too aroused.
Small:15
Medium:15
Large:9
3. I don’t play with my dog as a result
Small:2
Medium:1
Large:1
4. I use a game of “switch”.
Small:7
Medium:4
Large:8
5. I am currently undergoing a training plan given by a behaviourist.
Small:0
Medium:1
Large:1
6. I am currently training my dog with knowledge gained from online websites and or books.
Small:4
Medium:1
Large: 0
Hello Jo,
One very simple approach would be a chi-square test of independence, with one dimension being dog breed (3 levels) and the other dimension being the six options of owner response (so, a 3 x 6 contingency table).
However, the relatively small numbers of cases (of owners who did report play aggression) means that cell sizes (frequencies) will not be very large, so the statistical test won't have very much power.
Good luck with your work!
Hello Jo,
One very simple approach would be a chi-square test of independence, with one dimension being dog breed (3 levels) and the other dimension being the six options of owner response (so, a 3 x 6 contingency table).
However, the relatively small numbers of cases (of owners who did report play aggression) means that cell sizes (frequencies) will not be very large, so the statistical test won't have very much power.
Good luck with your work!
Agree with David Morse, and I am attaching a link which illustrate that how to perform cross tabulation and chi-square test in SPSS.
https://libguides.library.kent.edu/SPSS/ChiSquare
My Dear
when I am read your question I see that you have a 3 x 6 contingency table ( 2 factors) you can use foloowig statitistica test:
1- Chi-square to test inepenednet between your factors
2- Ratio test(propotion test ) by Z-statistic test
With Best Wishes
Hi,
setting aside questions about power, size of samples, etc, the main point, in my opinion, is that your various contingency tables are not independent because they collect answers of the same persons (owner or dog).
Instead of sparse analyses of different tables , one for each question, you might consider treat the whole set of answers as a single data set by a log-linear model (or alternatively by a Rasch model). This should give you a better undertanding of what your data mean.
You may Google these terms to get more information.
For a further help the data set would be mandatory.
I mean a data set with one line for each dog/owner (if they are more than 1 dog for some owners, neglect it as a start) and one column for each question, and yes or no in each cell.
Best wishes.
- Badges
- Science topic