frequency_filter: Residualizes child counts by regressing on adult counts for...

In rosemm/FrequencyFilter: The Frequency Filter

Description

Residualizes child word counts by regressing on counts for the main adult interlocotur for the same words in the same transcripts. The input data should already be word counts by speaker for each transcript or timepoint. By default, counts are log transformed before regressing.

Usage

frequency_filter(data_input, chi.count = "chi.count",
  adu.count = "adu.count", min.types = 200, log = TRUE, base = exp(1),
  drop.zeros = TRUE, outlier_cutoff = 1, debug = FALSE)

Arguments

data_input	the dataframe of counts in child and adult speech
chi.count	the name of the column with child counts
adu.count	the name of the column with adult counts
min.types	the minimum number of types necessary to run regression at a given timepoint or transcript. If set to 0, it will attempt to fit a regression for every timepoint/transcript regardless of how many words are available. Note that it is inadvisable to use residuals from regressions fit at very sparse timepoints.
log	log transform counts before regressing? Default is TRUE.
base	the base to use for log transformation
drop.zeros	should zero-count words (words that never occur in adult speech, or never in child speech) be excluded? Default is TRUE, which restricts analysis to counts of words that are observed in both child and adult speech.
outlier_cutoff	how should outliers (exceptionally high or low residuals) be defined? Default is 1 SD.
debug	turn on to display extra messages while the function runs, useful for identifying problem transcripts. Default is FALSE.

Value

The original input data, with new variables added on providing results from the regressions including residual values for each word.

References

Ibbotson, Hartman, and Nilsson Bj<c3><b6>rkenstam. (in prep). "Frequency Filter: An Open Access Tool for Analysing Language Development".

Examples

frequency_filter(data_pos)

rosemm/FrequencyFilter documentation built on May 29, 2019, 8:50 a.m.