weat_perm: WEAT Permutation Test

In conjugateprior/cbn: Tools and replication materials for Caliskan, Bryson, and Narayanan (2017)

Description

The statistic computed by this function is the mean cosine similarity of each x item to the a attributes minus the mean cosine to the b attributes, summed over x items subtracted for the same quantity computed for the y items. See the paper for details of the statistic, and the effect size.

Usage

weat_perm(items, vectors, x_name, y_name, a_name, b_name, b = 1000)

Arguments

items	information about the items, typically from cbn_get_items
vectors	a matrix of word vectors for all the study items, typically from cbn_get_item_vectors
x_name	the name of the target item condition, e.g. "Flowers" in WEAT 1
y_name	the name of the target item condition, e.g. "Insects" in WEAT 1
a_name	the name of the first condition, e.g. "Pleasant" in WEAT 1
b_name	the name of the second condition, e.g. "Unpleasant" in WEAT 1
b	number of bootstrap samples. Defaults to 1000.

Details

The p value is constructed by permuting the assignment of words to the x and y conditions. (The a and b attribute items are fixed.) The p value is the proportion of times the statistic computed on the permuted labels is greater than the value of the statistic that is observed.

Value

a data frame with first column the statistic, the second column the effect size, and the third column permutation test p value.

Examples

its <- cbn_get_items("WEAT", 1)
its_vecs <- cbn_get_item_vectors("WEAT", 1)
res <- weat_perm(its, its_vecs,
                 x_name = "Flowers", y_name = "Insects",
                 a_name = "Pleasant", b_name= "Unpleasant")
res

conjugateprior/cbn documentation built on June 26, 2019, 2:28 p.m.