1
$\begingroup$

I have a large survey data set which I am currently coding into SPSS. There are many times of questions, Likert scales, simple yes/no, simple agree/disagree and also rating scales. I can code them in but I need to do multiple linear regression analyses on the data and need to know if I should recode the yes/no and agree/disagree responses. They're currently in as categorical variables but will I be able to analyse it using MRA with all the questions being answered differently? Also, the categorical responses are 1=yes but one ordinal 5 point scale has the agree at the 5 and another has the agree at 1 so will I need to recode and reverse score anything?

$\endgroup$
  • $\begingroup$ This seems like a viable question to me. It doesn't have to be only a software-usage question. I'm voting to leave open. $\endgroup$ – gung Dec 2 '16 at 13:09
1
$\begingroup$

You do not need to reverse the direction of the negatively scored items for your multiple regression. What will happen is that the coefficient will be negative rather than positive (so, -2 instead of 2) but nothing else of importance will change. You might want to reverse the scoring if it makes it easier to talk about the results and that is sometimes a good idea. If you do choose to reverse some of them I would strongly recommend either (a) doing it into another variable (b) writing absolutely fool-proof documentation about what you did and storing it somewhere safe as if you forget later what you did you can waste a lot of time reconstructing it. As to your question about binary variables i assume SPSS handles that automatically but I do not use it so cannot reassure you.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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