Data Analysis

The 10 times rule fails badly — in one test it recommended 20 respondents where the true minimum was 599. Here are the exact G*Power steps, a lookup table, and the inverse square root cross-check.

Both — 0.85 for conceptually distinct constructs, 0.90 for similar ones. But since 2019 the test is whether the bootstrap upper bound is below the threshold, not whether the interval excludes 1.