Combining measures into 1 variable

[lsrms3]

I have 2 variables, one that measures (just for an examples) state aggression, and 1 that measures trait aggression. The state measure was responded to on a 6 point likert scale. The trait one was measured on a 5 point likert scale. Someone suggested that I combine these into one variable to use in analyses, suggesting that this would provide a more comprehensive look at it.

I'm having a hard time seeing how this would make any sense. Is this ever done? How does one go about it?

 

[kiton]

Indeed, it is plausible to combine such variables. To do so, I'd recommend you get antiquated with factor (and principal component) analysis.

 

[bugman]

Yes, or weight it if one measure has more importance over another. You would also need to standardise for the fact that one is measured on a five point scale while the other is one a five point scale.

 

[lsrms3]

Yes, or weight it if one measure has more importance over another. You would also need to standardise for the fact that one is measured on a five point scale while the other is one a five point scale.

Hi, can you explain how I would do that? Thanks!!

 

[freezo]

yes.. plz solve this problem

 

[rogojel]

Hi, can you explain how I would do that? Thanks!!

Hi,
the right way would be to use the principal component analysis but you might want to see if a simpler method would work. Transform both variables by standardising them x -> (x-Meanx)/Stdx then combine the two standardized variables into one (e.g. by taking the average). The reason you can do this is that by standardizing the variables you turned them into simple numbers without dimensions, so you are not combining apples with oranges. If you "know" that one variable is more important than the other you can use a weighted average as bugman suggested, but you should be able to argue for your choice of weights based on domain knowledge. E.g. "the choice of 0.2 0.8 leads to significant results" is NOT a good argument

regards

 

[sarabullock]

thnx.. it was helpful.

 

[Cabeceira]

Hi,
the right way would be to use the principal component analysis but you might want to see if a simpler method would work. Transform both variables by standardising them x -> (x-Meanx)/Stdx then combine the two standardized variables into one (e.g. by taking the average). The reason you can do this is that by standardizing the variables you turned them into simple numbers without dimensions, so you are not combining apples with oranges. If you "know" that one variable is more important than the other you can use a weighted average as bugman suggested, but you should be able to argue for your choice of weights based on domain knowledge. E.g. "the choice of 0.2 0.8 leads to significant results" is NOT a good argument

regards

hmmm. got it