Skip to contents

Simulated topic effects from an STM estimateEffect() object

Source: R/topic_effects.R
topic_effects.Rd

Tidy version of what plot.estimateEffect() draws: for each topic, the expected topic proportion over the values of a covariate, or the difference between two values, with simulation intervals:

  • "continuous": over a grid of n_points values between the minimum and the maximum of a numeric covariate (a year, a day);

  • "pointestimate": at each level of a categorical covariate (a type of document, a journal);

  • "difference": between cov_value1 and cov_value2.

By default, "continuous" for a numeric covariate and "pointestimate" otherwise.

Usage

topic_effects(
  effects,
  covariate,
  method = NULL,
  topics = NULL,
  cov_value1 = NULL,
  cov_value2 = NULL,
  moderator = NULL,
  moderator_value = NULL,
  n_points = 100L,
  n_sims = 500L,
  ci_level = 0.95,
  topic_labels = NULL,
  ...
)

Arguments

effects

An estimateEffect object (stm::estimateEffect()).

covariate

Character scalar, covariate name.

method

"continuous", "pointestimate" or "difference" (default: chosen from the type of the covariate).

topics

Integer vector of topics (default: those of effects).

cov_value1, cov_value2

Values compared when method = "difference".

moderator, moderator_value

Optional moderator (interaction) and its value.

n_points

Integer scalar, number of grid points of "continuous".

n_sims

Integer scalar, number of simulation draws.

ci_level

Numeric scalar in (0, 1).

topic_labels

Optional table with topic and label (e.g. from topic_labels()). NULL labels topics "Topic 1", "Topic 2", ...

...

Not used: an argument renamed in the development version of mintR gives an error naming its replacement.

Value

A data.table: method, topic, label, covariate, covariate_value, estimate, std_error, ci_level, ci_lower, ci_upper (and moderator, moderator_value when a moderator is given), sorted by topic and covariate value.

Details

Uses non-exported stm functions (produce_cmatrix, simBetas), the ones plot.estimateEffect() calls, so the results are those of the plot. The estimate is the mean of the simulations and the interval their quantiles.