\((1 - n/N)S_b^2/n\) with n the number of CLUSTERS, not
elements. 1000 units in 25 villages is 25 degrees of freedom, and
treating it as 1000 independent draws understates the standard
error by the square root of the design effect
\(1 + (\bar m - 1)\rho\). Both are computed, so the cost of
clustering is a number. Mirrors morie.fn.cluvar.
Usage
morie_cluster_variance(y, cluster, N = NULL)
Arguments
- y
element-level values.
- cluster
cluster identifiers.
- N
number of clusters in the population, for the fpc.
Value
list: mean, variance, se, n_clusters, n_elements,
mean_cluster_size, icc, deff, effective_n, naive_se,
se_inflation, method.
Examples
set.seed(1)
y <- rnorm(200) + rep(rnorm(20), each = 10)
morie_cluster_variance(y, rep(1:20, each = 10))$deff
#> [1] 6.169258
