The function nmathresh returns S3 objects of class thresh.
Objects of class thresh have the following components:
thresholdsA data frame with columns lo and hi
for the lower and upper thresholds, and lo.newkstar and
hi.newkstar for the new optimal (or rank-trt.rank) treatments
at each of the thresholds.
UThe threshold solutions matrix. One column for each data point \(m\), one row for each contrast \(d_{ab}\) (in ascending order). The elements \(U_{ab,m}\) describe the amount of adjustment to data point \(y_m\) required to reverse the relative ranking of treatments \(a\) and \(b\). This matrix is particularly useful for deriving thresholds for more complex decisions (e.g. bias-adjustment thresholds for a new treatment entering the top two, for any change in rank of the top three, etc.)
UkstarThe threshold solutions matrix limited to contrasts
involving \(k^*\). In other words, the rows of U corresponding to
contrasts of the form \(d_{ak^*}\) or \(d_{k^*a}\). Elements
\(U_{ak^*,m}\) of this matrix describe the amount of adjustment to data
point \(y_m\) required to make treatment \(a\) optimal (or
rank-trt.rank) over \(k^*\).
HThe influence matrix of the data on the basic treatment parameters. One column for each data point \(m\), one row for each basic treatment parameter \(d_k\). Elements \(H_{k,m}\) describe the influence of data point \(y_m\) on parameter \(d_k\). This matrix can be used to derive more complex thresholds (e.g. 2D thresholds for simultaneous adjustments to two data points, or thresholds for common adjustments to a group of data points).
kstarThe base-case optimal (or rank-trt.rank)
treatment \(k^*\).
callA list containing all the arguments defined in the
original call to nma_thresh.