Skip to contents

Pipeline stage: data intake and cleaning

Source: R/pipeline-intake.R
add_responses.Rd

Adds item responses to a pipeline, scores them with an answer key if one is given, and screens items and respondents before calibration. The checks (thresholds from mv_rules()):

Usage

add_responses(
  p,
  data,
  items = NULL,
  key = NULL,
  group = NULL,
  id = NULL,
  rt = NULL
)

Arguments

p

a pipeline from mv_pipeline().

data

a data frame (or matrix) of responses, possibly with other columns such as an ID or a grouping variable.

items

the item columns, by name or position. Defaults to every column that is not id, group or rt.

key

optional answer key: one correct option per item. Responses are then scored 1 (correct) or 0 (incorrect); missing responses stay missing.

group

optional grouping variable for DIF and group summaries: a column name of data or a vector with one value per respondent.

id

optional respondent IDs: a column name of data or a vector.

rt

optional response times: a matrix or data frame with one column per item (same order), or the names of those columns in data.

Value

The pipeline with the data and the intake flags added.

Details

  • zero_variance (block): every respondent gave the same answer, so the item cannot be calibrated.

  • possible_miskey (block): a 0/1 item with a very low proportion correct or a negative corrected item-total correlation, or a polytomous item with a negative correlation (possibly reverse-coded).

  • item_missing (review): an item with many missing responses.

  • person_missing (review): respondents who skipped most items.

  • straight_lining (review): respondents who gave the same answer to every item of a rating-scale (polytomous, unkeyed) test.

  • rapid_guessing (review): respondents with many responses faster than rapid_rt times the item's median response time (needs rt).

  • sparse_category (review): score categories with few responses.

  • small_group (info): groups too small for stable DIF results (fewer than 100 respondents).

Respondent-level checks raise one flag each, with the number of respondents as the target; the respondents themselves are listed in p$persons.

Examples

# \donttest{
p <- add_responses(mv_pipeline(), certexam, items = 3:34, key = certexam_key,
                   id = "id", group = "language")
flags(p)
#> # A tibble: 1 × 10
#>      id stage  check           target value      rule  level advice status note 
#>   <int> <chr>  <chr>           <chr>  <chr>      <chr> <chr> <chr>  <chr>  <chr>
#> 1     1 intake possible_miskey Q14    p = 0.18,… p < … block Check… open   ""   
# }