I'm looking to find information to recode a set of dichotomous response variables to preferable a 1-5 scale.

I have several response variables which when combined with PCA clearly depict certain strategies, ie.

1,0,0,0,0 = Accomodate / 1,1,0,0,0 = Accomodate / 0,0,1,0,0 = Retaliate / 0,0,1,1,0 = Retaliate / 0,0,0,0,1 = Reposition / 0,0,0,0,0 = Ignore

Discriminant analysis proves this model with 97% certainty.

Now, I would like to correlate these response strategies (seeing them as dependant variables) based on two scales I have created, being: Entry Barriers & Expected Retaliation (both measured on a 1-5 likert scale).

If I keep the response variables as 0-1 or even 1-5 I do not get any significant correlations, logically enough. However if I play around a little, ie. High Expected retaliation score translates to a 5 on Retaliation strategy, whilst low expected retaliation score translates to a 1-2 on retaliation strategy I get clear correlations with significance **.

I do not want to play around with data, I want to find a way to recode the binary variables to 1-5 scale variables for my correlations, this has led me to:

How to transform ordinal data from questionnaire into proper interval data?

Which leads me to IRT and Rasch model, which throws me off the bus, it is above my knowledge. I am guessing it essentialy assigns a sort of weighting to variables, but I cannot for the life of me figure it out.

This question is quintessential to my masters degree, hence I am looking for an answer desperately.

Any ideas? Or is this just impossible?


Your Answer

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

Browse other questions tagged or ask your own question.