2

I have a dataset with two variables: the response variable is the use of a fish species as a food resource, and the explanatory variable is the profession:

Person Profession Use of fish
John Fisherman Often
John Bus driver Often
Mary Fisherman Rarely
Joseph Joiner Occasionally
Martha Waitress Occasionally
Martha Fisherman Occasionally
Harry Fisherman Rarely
Thomas Waiter Often
Thomas Mason Often
Harold Fisherman Rarely
Harold Mason Rarely
Harold Carpenter Rarely
Christopher Fisherman Occasionally
Christopher Bus driver Occasionally

Note that Profession variable is repeated sometimes, since the same person may have two or more professions. Which statistical test should I use to compare Professions with regard to Use of fish or associate Profession with the Use? Could it be a Chi-square?

ttnphns
  • 51,648
  • 40
  • 253
  • 462
Arthur Filipe
  • 27
  • 6
  • 1
    One way to go with the investigation of the relation between Y=Use and X=Prof is to regress Y by X. Convert your data from "long" to "wide" structure: each Person is single row. Create a set of binary variables, each variable = one Prof, with values 1 (yes) or 0 (no). This is the predictors X. Perform multinomial regression of Y on X. For the analysis to be viable you must have consuderably more persons than professions, otherwise cut the number of professions first. – ttnphns May 16 '21 at 09:34
  • 1
    If it were not that one person in your data may have multiple professions then this analysis would in fact come to usual chi-square. – ttnphns May 16 '21 at 09:40
  • 1
    Please observe how I've edited your title and tags. I replaced "repeated measures" with "multiple response". It is a question where a respondent can select more than one variant of a categorical response. I consider this better than "repeated measures" because in repeated measures the measured value might repeat or be different. Here, same profession cannot repeat for the same respondent. If you don't like my changes, please feel free to roll back. – ttnphns May 16 '21 at 09:52
  • Thank you very much for your help! I really had thought to turn each profession into a binary variable, but I still didn't know if that would be adequate, or which test I could use. – Arthur Filipe May 16 '21 at 18:24
  • I found your changes in my question very pertinent, thank you very much for that too! – Arthur Filipe May 16 '21 at 18:28

0 Answers0