Package index • OpTop

Optimal number of topics

Parametric selection of the topic count that best describes the corpus.

optop_select() optimal_topic(): Select the optimal number of topics from a pool of topic models

Goodness-of-fit indices

Regression-style discrepancy indices and grid summaries for topic-model fits.

optop_index_se() optop_index_chisq() optop_index_deviance(): OpTop goodness-of-fit indices for a single topic model
optop_index_table(): OpTop goodness-of-fit indices table across a grid of LDA models

Harmonized comparisons across K

Fixed partition and baseline so all indices share a common support.

optop_make_partition(): Harmonized Rare-word Partition for OpTop indices
optop_make_baseline(): Global Corpus Baseline for OpTop Indices

Held-out validation and diagnostics

Score independent evaluation documents, select the smallest sufficient K, and test for structured residual bias.

optop_index_holdout(): Held-out goodness-of-fit indices with conditional inference
optop_gain_table(): Adjacent held-out gains and the epsilon-adequacy selection
optop_crossfit(): V-fold cross-fitted discrepancy indices
optop_moment_test(): Held-out moment-based specification tests

Utilities

Simulate corpora, shard very large ones, wrap external engines, and align inputs.

sim_dfm(): Simulate a document-feature matrix from an LDA specification
optop_corpus(): Sharded corpus for very large document sets
optop_warplda(): Wrap a text2vec WarpLDA fit for OpTop
optop_model(): Build an OpTop model from bare weight matrices
optop_align_dtm_to_models(): Align a document-term matrix to a grid of fitted models

Methods

Console summaries and plots for the objects OpTop returns.

print(<optop_theta_phi>) print(<optop_partition>) print(<optop_index>) print(<optop_index_table>) print(<optop_holdout>) print(<optop_moment_test>): Printing OpTop objects
plot(<optop_holdout>): Plot a held-out validation result
print(<optop_crossfit>) plot(<optop_crossfit>): Printing and plotting a cross-fitted result