generate: Generate stimuli

In JackEdTaylor/LexOPS: A Package and Shiny App for Generating Matched Stimuli

Generate stimuli

Description

Generates the stimuli from the data frame after it has been passed through split_by(), and optionally, control_for(). Will generate n items per condition. If <n items can be generated, will generate as many as possible given the experiment's design. Can be reproducible with seed argument.

Usage

generate(
  x,
  n = 20,
  match_null = "balanced",
  seed = NA,
  silent = FALSE,
  is_shiny = FALSE
)

Arguments

x	A LexOPS_pipeline object resulting from one of split_by(), control_for(), etc..
n	The number of strings per condition (default = 20). Set to "all" to generate as many as possible.
match_null	The condition that words should be matched to. Should be a string indicating condition (e.g. "A1_B2_C1"), or a string indicating one of the following options: "balanced" for randomly ordered null conditions with (as close to as possible) equal number of selections for each condition (default), "inclusive" to match each condition within the tolerances to every other condition, "first" for the lowest condition (e.g. "A1" or "A1_B1_C1_D1", etc.), "random" for randomly selected null condition each iteration.
seed	An integer specifying the random seed, allowing reproduction of exact stimuli lists. If NA, will not set the seed. Default is NA.
silent	Logical: should output to the console (via cat()) be suppressed? Default is FALSE.
is_shiny	Allows printing in a shiny context with shinyjs::html(). Outputs from the cat() function are stored in the div with id "gen_console". Default is FALSE.

Value

Returns the generated stimuli.

Examples

# Generate 20 words per condition, for design with 3 levels of syllables, controlled for frequency
lexops |>
  split_by(Syllables.CMU, 1:3 ~ 4:6 ~ 7:20) |>
  control_for(Zipf.SUBTLEX_UK, -0.2:0.2) |>
  generate(n = 20)

# Generate 2 levels of bigram probability, controlling for frequency and length
# (Note that the matching null is balanced across all stimuli)
# (Also note that the data is filtered by proportion known to be >75%)
lexops |>
  dplyr::filter(PK.Brysbaert >= .75) |>
  split_by(BG.SUBTLEX_UK, 0.001:0.003 ~ 0.009:0.011) |>
  control_for(Zipf.SUBTLEX_UK, -0.2:0.2) |>
  control_for(Length) |>
  generate(n = 1000, match_null = "balanced")

# Generate stimuli for a concreteness x valence (2 x 3) design
# (Note that abstract, neutral is set as the matching null)
lexops |>
  split_by(CNC.Brysbaert, 1:2 ~ 4:5) |>
  split_by(VAL.Warriner, 1:3 ~ 4.5:5.5 ~ 7:9) |>
  control_for(Zipf.SUBTLEX_UK, -0.25:0.25) |>
  control_for(Length) |>
  generate(n = 30, match_null = "A2_B2")

# As above but with inclusive tolerance
# (all words are within the specified tolerances relative to each other)
lexops |>
  split_by(CNC.Brysbaert, 1:2 ~ 4:5) |>
  split_by(VAL.Warriner, 1:3 ~ 4.5:5.5 ~ 7:9) |>
  control_for(Zipf.SUBTLEX_UK, -0.25:0.25) |>
  control_for(Length) |>
  generate(n = 30, match_null = "inclusive")

# Bypass non-standard evaluation
lexops |>
 split_by("Syllables.CMU", list(c(1, 3), c(4, 6), c(7, 20)), standard_eval = TRUE) |>
 control_for("Zipf.SUBTLEX_UK", c(-0.2, 0.2), standard_eval = TRUE) |>
 generate(n = 20)

# Create two levels of arousal, controlling for orthographic Levenshtein distance
library(stringdist)
lexops |>
 split_by(AROU.Warriner, 1:3 ~ 6:9) |>
 control_for_map(stringdist, string, 0:4, method="lv") |>
 generate(20)

# Create two levels of arousal, controlling for phonological similarity
library(stringdist)
lexops |>
 split_by(AROU.Warriner, 1:3 ~ 6:9) |>
 control_for_map(stringdist, eSpeak.br_1letter, 0:2, method="lv") |>
 generate(20)

# Create two levels of arousal, controlling for phonological Levenshtein distance, and rhyme
library(stringdist)
lexops |>
 split_by(AROU.Warriner, 1:3 ~ 6:9) |>
 control_for(Rhyme.eSpeak.br) |>
 control_for_map(stringdist, eSpeak.br_1letter, 0:2, method="lv") |>
 generate(20)

# A similar design to that above, but with 3 levels of valence, and inclusive matching
# Note that this will result in exactly the same result as above.
# A function that calculates something like Semantic Similarity will produce very different results.
library(stringdist)
lexops |>
 split_by(VAL.Warriner, 1:3 ~ 4.5:5.5 ~ 7:9) |>
 control_for(Rhyme.eSpeak.br) |>
 control_for_map(stringdist, eSpeak.br_1letter, 0:2, method="lv") |>
 generate(20, match_null = "inclusive")

JackEdTaylor/LexOPS documentation built on Oct. 11, 2024, 10:38 p.m.