Compute correlations

Compute correlations among pairs of words.

compute_corrs_at(data = NULL, token = NULL, feature = NULL,
  num_top_ngrams = 50, num_top_corrs = 50, adjust = FALSE,
  return_corrs = TRUE, return_words = FALSE, return_both = FALSE)

compute_corrs(..., token, feature)

Arguments

data	data.frame.
token	bare for NSE; character for SE. Name of column in data to use as item in widyr::pairwise_cor().
feature	bare for NSE; character for SE. Name of column in data to use as feature in widyr::pairwise_cor().
num_top_ngrams	numeric. Useful primarily to prevent widyr::pairwise_cor() from hanging up. If between 0 and 1, then interpreted as a percentage.
num_top_corrs	numeric. Useful primarily to limit input to a network visualation. does not hang up. If between 0 and 1, then interpreted as a percentage.
adjust	logical. Indicates whether or not to scale num_top_ngrams and num_top_corrs (if they are between 0 and 1) to the count of unique words. Is is likely that the input percentage values are specified with the number of words in mind.
return_corrs	logical. Whether to return just the correlations and token pairs, but not the token counts. This is the default option.
return_words	logical. Whether to return the counts used to compute the correlations. Useful when creating a network visualization.
return_both	logical. Useful when creating a network visualization so that words can be used as nodes and correlations can be used weights.
...	dots. Parameters to pass directly to visualize_time().

Value

data.frame.

Details

Calls widyr::pairwise_cor() internally.

See also

http://varianceexplained.org/r/seven-fav-packages/.