I'm struggling with my lab report. I have a questionnaire which contains Likert responses (1-7) as well as questions that require only yes or no answers. I want to see if there is a correlation between higher scores on the likert scale and responses to the binary answers to the other questions.
Specifically – the likert response is about anxiety levels. The binary (yes/no) answer is about negative consequences due to alcohol use. I want to just run the one ordinal likert item with all the negative consequence responses to see if there is a relationship. Can anyone help me?
Thanks so much for the fast comments. It definitely assures me that Spearman's Rho can be used but my main problem is the fact that i am mixing binary and ordinal variables together in the one correlation matrix. I have one ordinal variable (anxiety – 1-7 on likert scale) crossed with 13 dichotomous (0 or 1 representing yes or no) to answers ("have you ever been hospitalised due to your alcohol use?). I get statistically significant answers (<.05 and < .01 but the Correlation coefficients are tiny (i.e o.2) at best.
the hospitalisation is the DV an the anxiety levels the IV.
I just think i'm doing something wrong here. should i collapse the 13 binary variables in to one some how?
i can't see how SPSS will recognise the values between 1-7 when correlating with binary answers. i keep thinking it will just read that there is something in that box (ordinal) but not recognise the value (i.e 1-7) . but people keep saying that SPSS does in fact recognise this difference and account for it.
Best Answer
If you're comparing the correlation of two variables, just report the Spearman correlation coefficient. The Spearman correlation is recommended over Pearson correlation for this type of data: How to choose between Pearson and Spearman correlation?.
If you want to know how multiple variables impact the answer to one of the binary questions, do a logit or probit model. I've never used SPSS, but I'm sure there's a menu option for it. Select the binary answer as the dependent variable in the model.