Hi, I conducted a study on faking in personality tests, where the participants received two instructions (respond honestly vs. fake good) in a within-subjects design (oder was counter-balanced between-subjects).
I now want to compare different parameters regarding their classification sensitivity and specificity.
As I am using a within-subjects design I don't have a "state variable" (specifying whether the data is from the honest vs. faked condition) to compute a ROC analysis.
Any help in choosing the right procedure is highly appreciated. Thanks!
You can do it with gate recognition/face expression data, for which you can use SVM or Bayesian classifier.
you would need to analyze this using a causal inference framework.
why do you not have the treatment/intervention (your 'state') variable recorded?
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification