Is there a statistical equation to calculate sample size for accuracy, sensitivity, specificity, PPV and NPV for a diagnostic technique?
I think it depend on the study it self and these are varied in various studies
Mainly because of the incidence it self
It depends on your study design. Identifying the type of study design you want to do will help with the size of the sample needed to proceed.
Thanks Joseph
The study is to measure sensitivity and specificity of a new diagnostic tool against a gold standard.
Reliability sample size determination can be calculated by using Chii square distribution (or binomial distribution). However, Accuracy is tied with reliability; thus, Engagement is needed, such as Rel * Acc, gives us a new value. One must need to define reliability parameters before understanding accuracy measures. For example, a system will be 93% reliable if results are within a specific range (2 to 10 feet), but accuracy can be defined as 80% accurate if results are within 2-4 feet. Two different scenarios should be utilized to combine the reliability & Accuracy for sample determination. You can test 15 samples and only allow one failure (point estimate) needed to be (93% or higher) reliable. Now you have ten working systems (assuming two failures happened). Still, the accuracy requirement is 80% with 2-4 feet, now we need to determine actual sampling and allowable failures to show compliance combined (rel * Acc) requirements. In the above example, we are only allowing three failures and still can meet the requirements. Combined Rel*Acc = 0.93*0.8 =0.74.4
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence