 Solution

At stage 1, the maximum-likelihood estimators of the two normal arm means are their sample means. Their difference estimates the treatment effect:Each arm has $n_1$ patients and variance ${\sigma }^2/n_1$, so independence across arms givesThe quantity $I_1$ is the Fisher information for the mean difference when the common variance is known. The PDF's preliminary expression uses $n_i$ for the variance denominator, although $i$ indexes arms and $n_0$ is not defined. The stage-specific sample size is $n_j$; the calculation uses this intended reading, consistent with the specified $n_1,n_2$ per-arm recruitment and the later weighted estimator. Fresh stage-2 patients are independent of stage-1 patients in the usual trial model; the cumulative stage means themselves are not independent across stages.