We want to compare two sampling methods. All patients will go through a DNA test for diagnosis(positive or negative). Then samples are taken to identify the determine the type of bacteria and treatment. Current practice (method A) uses usual agar plates to take the samples and manually inoculate to culture them in the lab. Many times culture result come as negative for those with positive DNA results. There is another sampling device that takes samples in tubes and inoculate using robotic arms to culture in the lab (Method B). This method is easier for both patients and clinician and we think that samples will be better for culturing. For now we just want to show that Method B is as sensitive as Method A (when compare to DNA result). So we want to show that Method B is not worse that Method A in picking up bacteria. We only have a very rough estimate of the sensitivity of method A but I understand that the final comparison should be done by Mcnemar test since we are taking two sample using the two methods from each patient. I'm not sure on the sample size calculation. Do I need the sample size for Mcnemar test (for which I need the of diagonal estimates which we don't have!). Any help would be appreciated

  • $\begingroup$ So in summary, you have two methods A and B which you want to compare for differences, both methods take a sample from one patient. This is a classic paired t-test design. Why do you not have exact values for method A? Can you get them? Why do you need sample size? Are you trying to do power analysis? $\endgroup$ – user2974951 Jan 30 '19 at 15:15
  • $\begingroup$ DNA test confirms if the patient has infection. But the infection is not always picked up when sampling tissues- They take the sample , take it to the lab to culture but nothing is there. It comes as negative. For now aim is to show that method B is not worse than method A in detecting the infection (yes or no). We have a rough estimate of sensitivity for method A (compare to DNA test) from one clinic. I hope I'm using the medical terminology correctly. $\endgroup$ – Azita Jan 30 '19 at 18:13
  • $\begingroup$ So what exactly will you use to compare the two methods? A mean? A count of correct answers? What do you mean by sensitivity? What kind of variable is this? $\endgroup$ – user2974951 Jan 30 '19 at 22:02
  • $\begingroup$ Sensitivity or true positive- those positive by DNA are positive in the lab too $\endgroup$ – Azita Jan 31 '19 at 6:15
  • $\begingroup$ I am a little confused. McNemar's test works on paired nominal data, it's used to determine whether the row and column marginal frequencies are equal. Is this what you want? And if so, do you have the data for it? That is a value for each method A and B for each patient? If you do, then you have everything you need. So why are you mentioning sample size in your question? And diagonal estimates? $\endgroup$ – user2974951 Jan 31 '19 at 9:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.