Completes a matrix with missing ranks for the values maximally and minimally consistent with existing values using the Lim-Wolfe procedure
lwscreen(x,scrtest="lw.FriedmanTest")A matrix of ranks that may contain ties and NAs. Columns represent objects ranked and rows represent ranking methods.
What test to use to determine the maximally and minimally consistent imputed values.
The maximal and minimal statistics and p values for the list of completed rank matrices obtained.
lwscreen calls fillArows to impute the missing ranks in the matrix x. It then applies scrtest to all the matrices returned and finds the minimum and maximum values. See Lim and Wolfe (2002) for details of the algorithm.
The algorithm for finding the maximally consistent and inconsistent rank imputations is extremely computer intensive, creating large numbers of permuted matrices when tied ranksums or multiple rows with the same number of missing values are encountered. The APA election example in Lim and Wolfe (2002) is beyond the capability of the average PC in the present implementation.
Lim, D.H. & Wolfe, D.A. (2002) An efficient alternative to average ranks for testing with incomplete ranking data. Biometrical Journal, 43(2): 187-206.
# NOT RUN {
# The first example matrix from Lim and Wolfe (2002)
lwmat<-matrix(c(3,1,2,4,NA,2,1,NA,2,NA,1,NA),nrow=3,byrow=TRUE)
lwscreen(lwmat)
# }
Run the code above in your browser using DataLab