dict_reliability: Reliability analysis and PCA of a dictionary.
In PsychWordVec: Word Embedding Research Framework for Psychological Science

dict_reliability

R Documentation

Reliability analysis and PCA of a dictionary.

Description

Reliability analysis (Cronbach's \alpha and average cosine similarity) and Principal Component Analysis (PCA) of a dictionary, with visualization of cosine similarities between words (ordered by the first principal component loading). Note that Cronbach's \alpha can be misleading when the number of items/words is large.

Usage

dict_reliability(
  data,
  words = NULL,
  pattern = NULL,
  alpha = TRUE,
  sort = TRUE,
  plot = TRUE,
  ...
)

Arguments

`data`	A `wordvec` (data.table) or `embed` (matrix), see `data_wordvec_load`.
`words`	[Option 1] Character string(s).
`pattern`	[Option 2] Regular expression (see `str_subset`). If neither `words` nor `pattern` are specified (i.e., both are `NULL`), then all words in the data will be extracted.
`alpha`	Estimate the Cronbach's `\alpha`? Defaults to `TRUE`. Note that this can be misleading and time-consuming when the number of items/words is large.
`sort`	Sort items by the first principal component loading (PC1)? Defaults to `TRUE`.
`plot`	Visualize the cosine similarities? Defaults to `TRUE`.
`...`	Other parameters passed to `plot_similarity`.

Value

A list object of new class reliability:

alpha: Cronbach's \alpha
eigen: Eigen values from PCA
pca: PCA (only 1 principal component)
pca.rotation: PCA with varimax rotation (if potential principal components > 1)
items: Item statistics
cos.sim.mat: A matrix of cosine similarities of all word pairs
cos.sim: Lower triangular part of the matrix of cosine similarities

Download

Download pre-trained word vectors data (.RData): https://psychbruce.github.io/WordVector_RData.pdf

References

Nicolas, G., Bai, X., & Fiske, S. T. (2021). Comprehensive stereotype content dictionaries using a semi-automated method. European Journal of Social Psychology, 51(1), 178–196.

Examples

d = as_embed(demodata, normalize=TRUE)

dict = dict_expand(d, "king")
dict_reliability(d, dict)

dict.cn = dict_expand(d, "China", threshold=0.65)
dict_reliability(d, dict.cn)

dict_reliability(d, c(dict, dict.cn))
# low-loading items should be removed

PsychWordVec documentation built on April 4, 2025, 1:35 a.m.