sirt (version 4.1-15)

prmse.subscores.scales: Proportional Reduction of Mean Squared Error (PRMSE) for Subscale Scores

Description

This function estimates the proportional reduction of mean squared error (PRMSE) according to Haberman (Haberman 2008; Haberman, Sinharay & Puhan, 2008; see Meijer et al. 2017 for an overview).

Usage

prmse.subscores.scales(data, subscale)

Value

Matrix with columns corresponding to subscales

The symbol X denotes the subscale and Z

the whole scale (see also in the Examples section for the structure of this matrix).

Arguments

data

An \(N \times I\) data frame of item responses

subscale

Vector of labels corresponding to subscales

References

Haberman, S. J. (2008). When can subscores have value? Journal of Educational and Behavioral Statistics, 33, 204-229.

Haberman, S., Sinharay, S., & Puhan, G. (2008). Reporting subscores for institutions. British Journal of Mathematical and Statistical Psychology, 62, 79-95.

Meijer, R. R., Boeve, A. J., Tendeiro, J. N., Bosker, R. J., & Albers, C. J. (2017). The use of subscores in higher education: When is this useful?. Frontiers in Psychology | Educational Psychology, 8.

See Also

See the subscore package for computing subscores and the PRMSE measures, especially subscore::CTTsub.

Examples

#############################################################################
# EXAMPLE 1: PRMSE Reading data data.read
#############################################################################

data( data.read )
p1 <- sirt::prmse.subscores.scales(data=data.read,
         subscale=substring( colnames(data.read), 1,1 ) )
print( p1, digits=3 )
  ##                 A       B       C
  ## N         328.000 328.000 328.000
  ## nX          4.000   4.000   4.000
  ## M.X         2.616   2.811   3.253
  ## Var.X       1.381   1.059   1.107
  ## SD.X        1.175   1.029   1.052
  ## alpha.X     0.545   0.381   0.640
  ## [...]
  ## nZ         12.000  12.000  12.000
  ## M.Z         8.680   8.680   8.680
  ## Var.Z       5.668   5.668   5.668
  ## SD.Z        2.381   2.381   2.381
  ## alpha.Z     0.677   0.677   0.677
  ## [...]
  ## cor.TX_Z    0.799   0.835   0.684
  ## rmse.X      0.585   0.500   0.505
  ## rmse.Z      0.522   0.350   0.614
  ## rmse.XZ     0.495   0.350   0.478
  ## prmse.X     0.545   0.381   0.640
  ## prmse.Z     0.638   0.697   0.468
  ## prmse.XZ    0.674   0.697   0.677
#-> Scales A and B do not have lower RMSE,
#   but for scale C the RMSE is smaller than the RMSE of a
#   prediction based on a whole scale.