Skip to contents

Produces, for each item within each scale, the item-rest correlation, floor and ceiling proportions, and the item mean and standard deviation.

Usage

sframe_item_results(data, instrument, scales = NULL)

Arguments

data

A tibble or data.frame of responses.

instrument

An sframe object.

scales

Character vector or NULL. A subset of scale IDs to analyse. When NULL (default), all scales are included.

Value

An object of class sframe_item_report: a named list with one element per scale, each a list holding scale_id, label and diagnostics, a data frame with one row per item and columns item_id, mean, sd, item_rest_r, floor_prop, ceiling_prop and n_missing. The floor and ceiling shares are proportions. floor_pct and ceiling_pct, their names before 0.4.3, are kept as copies for 1 release. as.data.frame() stacks every scale's diagnostics into one table.

Details

Diagnostics use the scale's scoring orientation, so an item the scale reverse-codes is reversed first, as in score_scales() and sframe_reliability_results(). The item-rest correlation is the correlation between an item and the sum of the scale's other items. It is computed on respondents who answered every item in the scale, the same rows sframe_reliability_results() uses, and n_missing counts the item's own missing values in data.

Floor and ceiling are the proportions at the item's declared lowest and highest response, taken from its choice set, slider limits or rating maximum. They are NA for an item that declares no bounds, since the sample's own extremes say nothing about a floor or ceiling effect.

Examples

# \donttest{
demo <- sframe_demo("likert_scale")
ir <- sframe_item_results(demo$responses, demo$instrument)
print(ir)
# one scale's diagnostics
ir[[1]]$diagnostics
# every scale in one table
as.data.frame(ir)
# }