1,765 patients were given a positive/negative screening test for a disease.

N = 415 patients tested positive

N = 1,350 patients tested negative

Due to constraints, not all 1,765 patients were given the definitive disease diagnostic test; clinicians had to pick a sample size (95% CL, 5% MoE) from the positive and negative groups.

n = 200 patients tested positive

n = 300 patients tested negative

Does this mean that the True Positives (TP), FP, FN, and TN values have a +/- 5% MoE? So the 89% True Positives is 84 – 94%?

enter image description here

  • $\begingroup$ Hi all...any help would be greatly appreciated! $\endgroup$ – Harper Apr 5 '16 at 4:29
  • $\begingroup$ If this is a homework question, please add the self-study tag and read its wiki $\endgroup$ – Marquis de Carabas Apr 6 '16 at 8:20
  • $\begingroup$ Thanks for adding the tag. However, in addition to reading the tag wiki, you also needed to modify your question accordingly (as described in the guidelines there). The closure reason also outlines the problem. $\endgroup$ – Glen_b -Reinstate Monica Apr 11 '16 at 8:22

This seems like a strange and expensive way to run a study. Usually you would start with a group where you already know the disease state and run the screening test on them.

Here you start with the screening test and then sample that group to run the definitive test. This is not disqualifying...just expensive.

Anyway, you had to have presumed some hypothesized sensitivity and specificity to have run the power analysis on each proportion. These hypothesized values were almost certainly not realized. But even if they were, the sensitivity and specificity values are conditioned on the column totals, not the row totals (which is what the power analysis was done on). So the original values would not apply.


Not the answer you're looking for? Browse other questions tagged or ask your own question.