This piece assesses common ways states have combined data over the subgroup - grade level - year level to
increase effective sample size and statistical reliability.
We then approximate this distribution with a binomial distribution by fitting the mean and the variance of the binomial distribution to the bootstrapped distribution, yielding
an effective sample size of 18.4 individuals.
However, geophysical data are often autocorrelated, which has the effect of reducing
the effective sample size of the data.
With an overall sample size of 1005 children and an estimated intraclass correlation of 0.24 for the outcome of good SDQ score,
the effective sample size of ∼ 800 children would provide 80 % power, with a 5 % significance level, to detect differences of at least 12 % between the categories (and an outcome prevalence of at least 75 %) of study factors associated with a «good» SDQ score.