estimatr

Builds condition probability matrices for Horvitz-Thompson estimation from permutation matrix

permutations_to_condition_pr_mat(permutations)

Arguments

permutations

A matrix where the rows are units and the columns are different treatment permutations; treated units must be represented with a 1 and control units with a 0

Value

a numeric 2n*2n matrix of marginal and joint condition treatment probabilities to be passed to the condition_pr_mat argument of horvitz_thompson.

Details

This function takes a matrix of permutations, for example from the obtain_permutation_matrix function in randomizr or through simulation and returns a 2n*2n matrix that can be used to fully specify the design for horvitz_thompson estimation. You can read more about these matrices in the documentation for the declaration_to_condition_pr_mat function.

This is done by passing this matrix to the condition_pr_mat argument of

See also

declare_ra, declaration_to_condition_pr_mat

Examples

# Complete randomization perms <- replicate(1000, sample(rep(0:1, each = 50))) comp_pr_mat <- permutations_to_condition_pr_mat(perms) # Arbitrary randomization possible_treats <- cbind( c(1, 1, 0, 1, 0, 0, 0, 1, 1, 0), c(0, 1, 1, 0, 1, 1, 0, 1, 0, 1), c(1, 0, 1, 1, 1, 1, 1, 0, 0, 0) ) arb_pr_mat <- permutations_to_condition_pr_mat(possible_treats) # Simulating a column to be realized treatment z <- possible_treats[, sample(ncol(possible_treats), size = 1)] y <- rnorm(nrow(possible_treats)) horvitz_thompson(y ~ z, condition_pr_mat = arb_pr_mat)
#> Estimate Std. Error t value Pr(>|t|) CI Lower CI Upper DF #> z 0.09096484 0.5375002 0.1692369 0.8656103 -0.9625161 1.144446 NA