Variable Selection Method in Prediction Models: Application in Periodontology

Table 3

Diagnostic values for each logistic model.

	Sensitivity	Specificity	PPV	NPV
Logistic analysis (12 variables)	80.0%	73.1%	77.4%	76.0%
Logistic analysis	66.7%	61.5%	66.7%	61.5%
(4 variables: stearic, arachidonic, EPA, and DHA)
Logistic analysis with composite score	86.7%	69.2%	76.5%	81.8%
(3 variables: arachidonic, linoleic, and DHA)