Factor Analysis_Creating a Factor Scale vs a Summated Scale

[Adam Guerrero]

Hello,

I have conducted a factor analysis on 10 variables scaled 1 to 10 that assess customer perceptions using the commands

factor X6-X10 X12-X14 X16 X18, pcf
rotate, varimax

Since vars X6 and X13 loaded high on factor 4, results were used to construct a "summated scale" for the 4th factor as

generate NX13 = 10-X13
generate sf4 = (X6+NX13)/2

As you can see, I "reverse scored" X13 since its factor loading was negative.

Now, I want to create a factor score using predict as follows

factor X6-X10 X12-X14 X16 X18, pcf
rotate, varimax
predict  f4, regression

Here is my question. Do I have to first "reverse score" X13 before using predict to obtain factor scores?

Thanks so much,
Adam

[Clyde Schechter]

No, you don't have to reverse score X13 before using -predict- to obtain factor scores -predict- will handle the negatively loaded variable(s) appropriately.

[daniel klein]

Note that

predict f4, regression

will create the factor scores for the first (rotated) factor, not the fourth! Stata is clever but not that clever. You want

predict dump1 dump2 dump3 f4 , regression

to get the factor scores of the forth factor into f4.

Best
Daniel

[Adam Guerrero]

Thank you Clyde for your response. Also, Daniel, thanks for the tip - I did not know that.

Best,
Adam