Factor analysis
1. Find number of underlying dimensions
- Run principal component analysis
factor var1 var2 var3, pcf
- Apply rules for reducing dimensions:
- Kaiser criterion: eigenvalue > 1
- Check where screeplot tapers off by comparing with random data of similar size
fapara, pca reps(50)
- Check KMO statistics (measure of sampling adequacy)
estat kmo
0,5 > KMO > 0,7 => mediocre
0,7 > KMO > 0,8 => good
0,8 > KMO > 1,0 => great
2. Rotate: find factors that contain the same information but are easier to interpret
- Non-orthogonal (assumes underlying dimensions can be related)
rotate, promax blank(0.3)
estat common gives correlation between factors
- Orthogonal (assumes underlying dimensions are not related)
rotate, varimax
3. Calculate new variables
predict factor1
alpha q1-q9, gen(factor2)
Check if newly created variables correlate
corr factor1 factor2
1. Find number of underlying dimensions
- Run principal component analysis
factor var1 var2 var3, pcf
- Apply rules for reducing dimensions:
- Kaiser criterion: eigenvalue > 1
- Check where screeplot tapers off by comparing with random data of similar size
fapara, pca reps(50)
- Check KMO statistics (measure of sampling adequacy)
estat kmo
0,5 > KMO > 0,7 => mediocre
0,7 > KMO > 0,8 => good
0,8 > KMO > 1,0 => great
2. Rotate: find factors that contain the same information but are easier to interpret
- Non-orthogonal (assumes underlying dimensions can be related)
rotate, promax blank(0.3)
estat common gives correlation between factors
- Orthogonal (assumes underlying dimensions are not related)
rotate, varimax
3. Calculate new variables
predict factor1
alpha q1-q9, gen(factor2)
Check if newly created variables correlate
corr factor1 factor2
