Estimate comparison

This is a comparison of A/B testing with a fixed samplesize test versus a sequential GLR test (via SeGLiR). Click "Simulate" to generate data from the given proportions p1 and p2 and simulate outcomes and MLE estimates from the test. Click "Calculate bias-adjusted MLE" to calculate the Whitehead bias-adjusted MLE estimate for the sequential GLR test - note that this may take a while to calculate, especially if the true difference is very small.

Both tests are "comparison of proportions" tests done at α-level 0.05, β-level 0.10, where the null-hypothesis is that p1 = p2, while the alternative hypothesis is that p1 p2. We also specify that we are interested in detecting any difference between p0 and p1 larger than 0.01. Note that the fixed sample-size test usually gives more precise estimates due to usually having more samples. The sequential GLR also gives biased estimates, which can be corrected by the Whitehead bias-adjusted MLE.

For more information about the sequential GLR test take a look at the reference of SeGLiR.

Fork me on GitHub