predict_style: Simulate Data Distorted by Response Style and Calculate, for...

In hplieninger/stylesim: Simulate (and Analyse) Data Distorted by Response Styles

Description

This function repeatedly simulates data that are distorted by response style. It makes use of the scale scores \bar{X}—which are influenced by both the attribute and response style—as well as the true person parameters θ_1, that is, the persons' standing on the attribute, which is unaffected by response style. These two scores are used to classify each individual at certain cutoff values as either above (i.e., positive) or below (i.e., negative) the cutoff. Subsequently, the true positive rate (TPR), the false positive rate (FPR), the precision, and the accuracy are calculated at every cutoff value.

Usage

predict_style(reps = 100, n = 200, items = 10, categ = 5, ndimc = 1,
  style = NULL, reversed = 1/3, mu.s = 0, var.s, sig = NULL,
  emp = TRUE, cutoff, ...)

Arguments

reps	Desired number of replications
n	Numeric. Desired number of observations/persons.
items	Numeric. Desired number of items/stimuli. If ndimc > 1, this specifies the number of items per dimension.
categ	Numeric. Desired number of categories per item.
ndimc	Numeric. Desired number of content-related latent variables (irrespective of number of style-related latent variables).
style	Parameter to specify which response style(s) influence the data, can be either numeric or character. Users may choose one or more among "ERS1" (e.g., 1 / 0 / 0 / 0 / 1), "ERS2" (e.g., 2 / 1 / 0 / 1 / 2), "ARS" (e.g., 0 / 0 / 0 / 1 / 1), "ADRS" (e.g, -1 / -1 / 0 / 1 / 1), and "MRS" (e.g., 0 / 0 / 1 / 0 / 0). Alternatively, a user-specified vector of weights can be employed. Can also be NULL indicating complete abscence of response styles.
reversed	Numeric. Indicates the number of reverse-coded items. Can be either a ratio between 0 and 1 indicating the ratio of reverse-coded items or an integer between 0 and items) indicating the number of reverse-coded items.
mu.s	Numeric. The mean(s) of the response style distribution.
var.s	Numeric. The variance(s) of the response style distribution.
sig	Numeric matrix. The variance-covariance matrix of the multivariate distribution of thetas. If non-NULL, this overrides var.s.
emp	Logical. If true, mu.s and var.s/sig specify the empirical not population mean and covariance matrix.
cutoff	Optional numeric vector. Desired cutoff- (alpha-) values at which to calculate TPR, FPR, etc; probably a sequence, for example from .95 to .05 (must be within the open interval of (0, 1))
...	Other parameters passed to sim_style_data.

Value

alpha	The cutoff-/alpha-values
TPR	True positive rate
FPR	False positive rate
PREC	Precision
ACC	Accuracy
	Further elements containing the input parameters

See Also

The replicated function sim_style_data.

hplieninger/stylesim documentation built on May 17, 2019, 4:54 p.m.