I have a data set that is represented below, but much longer. How can I find if there is a statistical relationship between business category and business score?

Business Name     Business Category        Business Score
Business1         Shopping                      1
Business2         Eating                        1
Business3         Shopping                      2
Business4         Shopping                      2
Business5         Eating                        1
Business6         Manicure                      3
Business7         Haircut                       2
Business8         Haircut                       3
Business9         Eating                        3
Business10        Manicure                      3
Business11        Haircut                       3
Business12        Eating                        1
Business13        Shopping                      2
Business14        Eating                        3
Business15        Manicure                      1
Business16        Haircut                       3
Business17        Haircut                       2
Business18        Shopping                      3
Business19        Eating                        2
Business20        Manicure                      3

2 Answers 2


Given that business is categorical, you can't use correlation.

Business score looks like it is a 1-5 rating or something like that. If that is the case, I would recommend ordinal logistic regression In SAS use PROC LOGISTIC. In R there is polr. The dependent variable would be "score" and the independent variable "type of business".

  • $\begingroup$ The OP appears to use the term "correlation" in the nontechnical sense of a "relationship." A constructive way to interpret that is that it's a request to identify significant differences in score among business categories. $\endgroup$
    – whuber
    Mar 12, 2014 at 16:14
  • $\begingroup$ Hi, I guess I should not have used the term correlation. I am only looking for a significant relationship between business category and business score. Does the above answer still satisfy this? $\endgroup$
    – metersk
    Mar 12, 2014 at 16:17
  • $\begingroup$ Yes it does. Do you know how to do it? $\endgroup$
    – Peter Flom
    Mar 12, 2014 at 17:15
  • $\begingroup$ @PeterFlom I believe so? I imported the arm package and the data and did this: x <- polr(Business.Category ~ Score, data = data). Then I did summary(x) and got this output, pastebin.com/sBGg1xEG, although I'm not really sure what to make of it. $\endgroup$
    – metersk
    Mar 12, 2014 at 18:34
  • $\begingroup$ @PeterFlom Also, are there any other descriptive statistics I can show on x that would show a significant relationship, maybe p-value or r-squared? $\endgroup$
    – metersk
    Mar 12, 2014 at 18:50

I think you can try a simple Chi-square test. Make a frequency table of "Business Category" vs "Business Score" and then go for a chi square test. I don't have R in this PC. Later I will post some code.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.