diagnostic test accuracy measures

[Eve Martin]

I have a dataset that measures cancer (d), PSA level (psa), and test positive/negative (tp). A PSA>4 correlates to a positive test result, and all other PSA values correlate to a negative test result. cancer (d) is a binary qualitative variable, either yes or no.

I am wondering how to calculate diagnostic test accuracy measures to fill out this table

		criterion standard	criterion standard
		Abnormal	Normal	Total
test result	Positive (PSA>4)	TP	FP	T+
test result	Negative (PSA<4)	FN	TN	T-
	total	D+	D-	n

TP=true positive
FP=false positive
FN=false negative
TN=true negative

[Clyde Schechter]

label define pos_psa 0 "Negative" 1 "Positive"
gen byte pos_psa:pos_psa = (psa > 4 & !missing(psa))

tab pos_psa d

And if you add the -col- option to the -tab- command you will get the sensitivity and specificity. And if you add the -row- option to the -tab- command you will get the positive and negative predictive values. You didn't ask for those, but I imagine getting those is your next step.