For example, if I'm screening for cancer and test 100 people in group A then 100 people in group B and get the test fail rate of 4% in group A and 6% in group B, what statistical test can I use to compare the two fail rates?
As a follow up question, if I'm calculating the sample size I need to screen in Group A and Group B to be able to find a difference between the two, how can I calculate this? If the expected fail rate for group A is 4%, and 6% for group B and the standard deviation in the fail rate for groups is 2% (and trying to get a power of 0.8 with alpha 0.05 and a sampling ratio of 1).
Chi-squared or logistic regression.
I'm not sure about the power analysis for a chi-squared test, but I just ran some simulations of different sample sizes with your expected fail rate for groups A and B, which suggested you would need between 900 and 1000 per group for that difference to become significant at alpha=.05.
Chi-squared or logistic regression.
I'm not sure about the power analysis for a chi-squared test, but I just ran some simulations of different sample sizes with your expected fail rate for groups A and B, which suggested you would need between 900 and 1000 per group for that difference to become significant at alpha=.05.
I must agree. A logistic regression modeling counts (percentage usually convert to counts) may be well utilized when these scenarios occur.
- Badges
- Science topic