I am stuck with a problem and hope you people can help me out.

In simple words, I have been researching about the impact of input factors on the output. My input factors consists of 8 variables of which 6 are in the form of dichotomous data i.e. Yes/No and 2 are in the form of Likert scale data 1 to 8. My output is the revenue generated which is in the terms of monetary value (For eg. 3.45 Million). I have a data set of close to 80000 values. Can anyone of you suggest how to apply factorial analysis using SPSS and what type of analysis should be applied??

  • $\begingroup$ The question is unclear, especially what expression "output factor" means $\endgroup$ – ttnphns Jul 3 '17 at 7:18
  • $\begingroup$ Output factor means the result obtained from employing the input factors. For eg. The score in your tests is the direct output of the number of correct questions you answer. $\endgroup$ – Shantanu Roy Jul 3 '17 at 7:19
  • $\begingroup$ Let me make it more clear i want to interpret whether training (One of the Input factors) my employees has resulted in any increase in their performance(output). $\endgroup$ – Shantanu Roy Jul 3 '17 at 7:22
  • $\begingroup$ In statistics and in science in general, word "factor" implies something that influences. Something that is a cause or a moderator, in conceptual view. Not the result or output. That's why "output factor" is strange to me unless you define what that factor influences in your study. Note also that I erased tag [factor-analysis] because I didn't see how your question is connected to specifically that type of analysis. $\endgroup$ – ttnphns Jul 3 '17 at 7:27
  • $\begingroup$ Ok. So can you just tell me how can i incorporate my input factors which are a combination of Dichotomous data, Likert scale type data and Rank order type data in SPSS so that i can analyze its effect on my output? $\endgroup$ – Shantanu Roy Jul 3 '17 at 10:49

It sounds like you want some kind of regression, with revenue (or maybe log(revenue)) as the dependent variable and the others as independent variables.

Standard methods don't deal well with ordinal independent variables (such as Likert scaled items) but a relatively little used method called optimal scaling can deal with them quite well.

| cite | improve this answer | |
  • $\begingroup$ Thanks Peter. Can you brief me about optimal scaling? $\endgroup$ – Shantanu Roy Jul 3 '17 at 10:59
  • $\begingroup$ Essentially, it's a method of coding variables so that they "work best" for the regression. There are coding schemes for nominal, ordinal, and continuous variables. I wrote a paper about doing this in SAS: Alternative methods of regression when OLS is not right . You can also do this in R. $\endgroup$ – Peter Flom Jul 3 '17 at 11:03

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.