Need help with analysing a survey on minitab

#1
Hi all,

I'm working on a project for uni where we are are required to use minitab (with very little direction or help) and I am stuck on the basics of analysis. Apologies if I'm not talking the stats lingo, but I am completely new to this!

The questionnaire consists of 26 questions and has 178 total responses from cat owners. 8 questions consist of yes/no answers, open questions and answers where male/female are the only options. The remaining 16 questions/statements are based on Likert scale scores (1= Strongly Disagree, 2= Disagree, 3=Neither/Not Applicable, 4= Agree and 5= Strongly Agree) - for example "my cat hides from strangers" or "I consider my cat to be aggressive".

I am trying to see whether there are any relationships between kitten experience and behaviours later on in life (Adulthood). For example, 62 out of my 178 responses have answered "YES" in response to the statement "My kitten was weaned before 8 weeks" (below the crucial learning age of the kitten- this is a CAUSE question) and I want to find out whether there is a relationship between this statement and anxiety behaviours from each response such as "hiding from strangers" from the Likert score.

Each CAUSE question is to be tested against 8 different VARIABLE behaviours.

All of the data has been tested for normailty. Some data are not normal and some are normal distribution, so I have a potential mixture of parametric/non-parametric tests.

Please could anyone enlighten me to how I should arrange my data for testing and which test I should do for analysis to find a relationship? Below is a visual representation of how my data is currently arranged. Each row is the response of one participant.
1588862436239.png




Any help will be highly appreciated

Many thanks
Stace
 

Miner

TS Contributor
#2
First of all, don't expect normality from ordinal (Likert scale) data. Normality is for ratio and interval data, not ordinal. I would recommend starting with exploratory data analysis. Use graphical tools like a dot plot to stratify your data by groups (CAUSE questions). If there is a relationship of practical value, you should be able to see it graphically. You might try a scatter plot of the ordinal variables against age (be sure to enable jitter, so the plotted points don't fall on top of each other). The selection of which test to use will depend very much on what you find.

1588868454802.png
 

noetsi

Fortran must die
#3
I knew ordinal predictors were always linear. I did not know that non-normality was not an issue with them.