Using the fitted responses (probabilities) to the difference scale, new responses are generated which permit new bootstrap replications of estimated scales to be generated. The mean scale is useful for evaluating bias and the standard deviation for estimating standard errors of the scale values.
boot.mlds(x, nsim, no.warn = TRUE, ...)
boot.mlbs(x, nsim, no.warn = TRUE, ...)A list of 4 elements of class ‘mlds.bt’:
A \(p + 1\) x \(N\) matrix of the bootstrap samples of the scale, where p is the number of scale values and N is the number of simulations. The extra value is the bootstrapped \(sigma\) value.
A vector of length \(p + 1\) giving the mean of the bootstrap scales.
A vector of length \(p + 1\) giving the standard deviation of the boostrap scales.
The number of bootstrap simulations.
an object of class ‘mlds’ or ‘mlbs’ depending on which function is called.
an integer, the number of simulations.
logical indicating when TRUE (default) to suppress warnings from glm
Additional options passed along to the function mlds.
Kenneth Knoblauch and Laurence T. Maloney
Either the scale values (from ‘glm’ method) or the scale values and \(\sigma\) (from ‘optim’ method) permit the fitted probabilities to be estimated. These are used to generate new responses to the quadruples using rbinom. The new responses are then used with mlds to estimate a bootstrapped scale. This is repeated \(N\) times and stored in the output with the mean and standard deviation of the bootstrapped scales.
Maloney, L. T. and Yang, J. N. (2003). Maximum likelihood difference scaling. Journal of Vision, 3(8):5, 573--585, tools:::Rd_expr_doi("10.1167/3.8.5").
Knoblauch, K. and Maloney, L. T. (2008) MLDS: Maximum likelihood difference scaling in R. Journal of Statistical Software, 25:2, 1--26, tools:::Rd_expr_doi("10.18637/jss.v025.i02").
mlds
data(kk1)
kk1.mlds <- mlds(kk1)
#nsim should be near 10,000 for stability,
# but this will take a little time
boot.mlds(kk1.mlds, 100)
Run the code above in your browser using DataLab