I am a newbie when it comes to SPSS, does anyone know which test I need to use for my analysis, and how to do this?

#1
Hi, I'm new on this forum and new to statistics and SPSS as well (it is not my strong suit, unfortunately, but I have to do this for a project). I am trying to investigate the relationship between the frequency of cannabis usage and the respective alcohol, tobacco and illicit drug use in patients with a headache disorder. All the data I have is fictional, but in my assignment is stated that the data is acquired by a questionnaire that the patients have filled in.

I have divided my dataset into 7 different groups depending on the self-reported frequency of their cannabis usage ('never', 'incidental', 'sometimes', 'regularly' and 'past, incidental', 'past, sometimes', and 'past, regularly').

The variables 'alcohol use', 'tobacco use' and 'illicit drug use' are all categorical variables as well, as the patients could fill in 'yes', 'never', 'sometimes', 'quit' or 'NA'.

Now, I have found a way to get to know how many patients for example consumed alcohol 'sometimes' or 'never' in all of the 7 different groups by splitting the file based on cannabis usage and using the 'frequencies' option for the variables i wanted.
However, I am looking for a way to see if there is any significant difference between the distributions of alcohol usage (as well as tobacco and illicit drug usage) between these different groups.
Does anyone have any suggestions on how I can best approach this? I also think I need to find a way to convert these numbers into percentages as the 7 different groups are all different in sample size, so absolute numbers won't be useful in the analysis, right?

I have tried performing a chi-squared test as I thought it would be best suited for this type of analysis, however, it does not show the results I am looking for. I have put my outcome in the attachments.

I don't know if I have explained my issue well enough as English is not my first language, so if there is anything that is unclear please do say so.
Thanks in advance :)
SPSS outcome.png
 

Karabiner

TS Contributor
#2
I have divided my dataset into 7 different groups depending on the self-reported frequency of their cannabis usage ('never', 'incidental', 'sometimes', 'regularly' and 'past, incidental', 'past, sometimes', and 'past, regularly').

The variables 'alcohol use', 'tobacco use' and 'illicit drug use' are all categorical variables as well, as the patients could fill in 'yes', 'never', 'sometimes', 'quit' or 'NA'.
Wel, AFAICS you do not have a problem concerning the software SPSS, but a statistical problem.

The respsonses you analyse are not purely categorical (like hair coulour: "grey", "blonde",
"brown", black" etc.), but there is some ordering information in them, which you could
use (you can treat ordinal scaled responses as categorical, but it wastes information and
often produces a too complex pattern of results). You could code the levels as e.g. 1=never,
2=past/incidental etc., (or whatever makes sense), analyse the data using the Spearman
correlation. Spearman gives you the degree of association between the variables
in the sample. In addition it is tested automatically for statistical significance in SPSS.

With kind regards

Karabiner
 
Last edited: