Reliability of test

Usage

RItif(dfin, lo = -5, hi = 5, samplePSI = FALSE, cutoff = 3.33, model = "PCM")

Arguments

dfin

Dataframe with item data only

lo

Lower limit of x axis (default = -5)

hi

Upper limit of x axis (default = 5)

samplePSI

Adds information about sample characteristics

cutoff

Caption text will generate information relative to this TIF value

model

Defaults to "PCM", use "RM" for dichotomous data

Details

NOTE: TIF is not reliable with small numbers of items, and should not be used as your primary indication of reliability (see doi: 10.1111/bmsp.12033). Please use RIreliability() and RIrelRep() instead.

Test information shows the reliability curve of the test (not the sample).

Use option samplePSI = TRUE to add graphical and written representation of the current sample's theta mean/SD, test information (TIF) mean/SD, and Person Separation Index (PSI). According to Wright & Stone (1999), PSI is calculated as \(\frac{\mathrm{SSD}-\mathrm{MSE}}{\mathrm{SSD}}\), see ?eRm::SepRel for details. According to Embretson & Reise (2000), PSI = 1 - SEM^2, and TIF = 1/SEM^2, and the values reported in this function are based on sample average SEM.

For reference: TIF 2.5 corresponds to PSI 0.6 TIF 3.33 -> PSI 0.7 TIF 5 -> PSI 0.8 TIF 10 -> PSI 0.9

References

Milanzi, et al. (2015). Reliability measures in item response theory: Manifest versus latent correlation functions. doi:10.1111/bmsp.12033