Does direction of data matter on applying factor analysis like PCA? I have dataset of water quality parameters. Among of those parameter, the DO (Dissolve Oxygen) is inverse to others, i.e. higher DO means better quality unlike higher BOD5 means worse. Do I have to reverse direction of DO in order to apply factor analysis?
Thank you
Nice question . Have you carried out partial regression analysis. ?
If not done , keep DO as top order parameter...
Hello Hong,
No, you needn't reverse the sign ("polarity") of variables in order to evaluate their affiliations within a larger set. The key is to note (as your query suggests you are well aware) the meaning of positive and negative relationships in the context of the variables in your set, so as not to misinterpret the results.
One small note: PCA (principal components analysis) isn't quite the same thing as factor analysis, though they are often conflated in published studies (as if interchangeable). PCA assumes all observed variation is common variation, and that there is no unique or specific variation (unshared) in the data set, nor is there any error variation. When a PCA solution is compared to a common factor analysis solution, the estimated loadings for PCA tend to be biased higher (which could affect interpretation of salience).
Good luck with your work.
It is not really necessary even though variables have different units and scales within a dataset since at the beginning step all variables are standardization
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts