ShinyItemAnalysis (version 1.3.4)

ItemAnalysis: Item Analysis

Description

ItemAnalysis function computes various traditional item analysis indices including difficulty, discrimination and item validity. For ordinal items the difficulty and discrimination indices take into account minimal item score as well as range.

Usage

ItemAnalysis(data, y = NULL, k = 3, l = 1, u = 3,
maxscore, minscore, cutscore, add.bin = FALSE)

Arguments

data

matrix or data.frame of items to be examined. Rows represent respondents, columns reperesent items.

y

vector of criterion values.

k

numeric: number of groups to which may be data.frame x divided by the total score. Default value is 3. See Details.

l

numeric: lower group. Default value is 1. See Details.

u

numeric: upper group. Default value is 3. See Details.

maxscore

numeric or vector: maximal score in ordinal items. If missing, vector of obtained maximal scores is imputed. See Details.

minscore

numeric or vector: minimal score in ordinal items. If missing, vector of obtained minimal scores is imputed. See Details.

cutscore

numeric or vector: cut score used for binarization of ordinal data. If missing, vector of maximal scores is imputed. See Details.

add.bin

logical: If TRUE, indices are printed also for binarized data. See Details.

Value

ItemAnalysis function computes various traditional item analysis indices. Output is a data.frame with following columns:

diff

average score of the item divided by its range

avgSCore

average score of the item

SD

standard deviation of the item score

SDbin

standard deviation of the item score for binarized data

correct

proportion of correct answers

min

minimal score specified in minscore; if not provided, observed minimal score

max

maximal score specified in maxscore; if not provided, observed maximal score

obtMin

observed minimal score

obtMax

observed maximal score

cutScore

cut-score specified in cutscore

gULI

generalized ULI

gULIbin

generalized ULI for binarized data

ULI

discrimination with ULI using the usual parameters (3 groups, comparing 1st and 3rd)

ULIbin

discrimination with ULI using the usual parameters for binarized data

RIT

item-total correlation (correlation between item score and overall test score)

RITbin

item-total correlation for binarized data

RIR

item-rest correlation (correlation between item score and overall test score without the given item)

RIRbin

item-rest correlation for binarized data

itemCritCor

correlation between item score and criterion y

itemCritCorBin

correlation between item score and criterion y for binarized data

valInd

item validity index calculated as cor(item, y)*sqrt(((N-1)/N)*var(item)), see Allen & Yen (1979), Ch.6.4

valIndBin

item validity index for binarized data

rel

item reliability index calculated as cor(item, test)*sqrt(((N-1)/N)*var(item)), see Allen & Yen (1979), Ch.6.4

relBin

item reliability index for binarized data

relDrop

item reliability index 'drop' (scored without item)

relDropBin

item reliability index 'drop' (scored without item) for binarized data

alphaDrop

Cronbach's alpha without given item

alphaDropBin

Cronbach's alpha without given item, for binarized data

missedPerc

proportion of missed responses on the particular item

notReachPerc

proportion of respondents that did not reached the item nor the subsequent ones, see recode_nr function for further details

With add.bin == TRUE, indices based on binarized data set are also provided and marked with bin suffix.

Details

For ordinal items the difficulty and discrimination indices take into account minimal item score as well as range.

For calculation of discimination ULI index, it is possible to specify the number of groups k, and which two groups l and u are to be compared.

In ordinal items, difficulty is calculated as difference of average score divided by range (maximal possible score maxscore minus minimal possible score minscore).

If add.bin is set to TRUE, item analysis of binarized data is included in the output table. In such a case, cutscore is used for binarization. When binarizing the data, values greater or equal to cut-score are set to 1, other values are set to 0.

References

Martinkova, P., Stepanek, L., Drabinova, A., Houdek, J., Vejrazka, M., & Stuka, C. (2017). Semi-real-time analyses of item characteristics for medical school admission tests. In: Proceedings of the 2017 Federated Conference on Computer Science and Information Systems. https://doi.org/10.15439/2017F380

Allen, M. J. & Yen, W. M. (1979). Introduction to measurement theory. Monterey, CA: Brooks/Cole.

See Also

DDplot, gDiscrim, recode_nr

Examples

# NOT RUN {
# loading 100-item medical admission test data sets
data(dataMedical, dataMedicalgraded)
# binary data set
dataBin <- dataMedical[, 1:100]
# ordinal data set
dataOrd <- dataMedicalgraded[, 1:100]
# study success is the same for both data sets
StudySuccess <- dataMedical[, 102]

# item analysis for binary data
head(ItemAnalysis(dataBin))
# item analysis for binary data using also study success
head(ItemAnalysis(dataBin, y = StudySuccess))

# item analysis for binary data
head(ItemAnalysis(dataOrd))
# item analysis for binary data using also study success
head(ItemAnalysis(dataOrd, y = StudySuccess))
# including also item analysis for binarized data
head(ItemAnalysis(dataOrd,
  y = StudySuccess, k = 5, l = 4, u = 5,
  maxscore = 4, minscore = 0, cutscore = 4, add.bin = TRUE
))
# }