Multiple Comparisons Methods
Choose a method for critical point calculation from the dropdown list. The following options are available:
best
The smallest critical point, choosing from all the valid methods, is provided.
best.fast
The smallest critical point, choosing from all the valid methods except Simulation, is provided. This is the default.
Bonferroni
If a total of m bounds will be computed (counting each interval as 2 bounds), the critical point by this method will be qt(1-alpha/m, df.residual).
Dunnett
If the linear combinations specify all differences between one element of bvec or column of lmat (the control) and several others, the critical point for comparisons-with-control intervals or bounds will be used, if valid. The validity condition requires that the covariance matrix of the treatment-control differences to be equivalent to that of a one-factor model (allowing unequal sample sizes). You can override the validity check by clearing the Validity Check box.
Fisher.lsd
The critical point for two-sided intervals is the upper-alpha/2 Student's-t value, qt(1-alpha/2, df.residual); for bounds, the upper-alpha point. This is the only method available if Error Type is comparison-wise, and is not available if Error Type is family-wise.
Scheffe
If the rank of the covariance matrix of the estimators of the linear combinations of interest is Srank, the critical point will be sqrt(Srank*qf(1-alpha, Srank, df.residual)).
Sidak
If Bounds is both, the critical point will be the upper-alpha quantile from the maximum absolute value of k "uncorrelated" multivariate-t random variables. If Bounds is upper or lower, the critical point will be the corresponding quantile of the maximum without taking absolute values, if the estimators specified by lmat and /or comparisons are uncorrelated (or if Validity Check is cleared).
Simulation
An approximate critical point is generated using the simulation-based method. Under the default simulation size, the critical point generated will give a family-wise error rate within 10% of the specified alpha (with 99% confidence).
Tukey
If the linear combinations specify all pairwise differences between several quantities, specifying this method requests the critical point to be the Tukey studentized-range quantile scaled by sqrt(2). If more than three quantities are to be compared, validity of the Tukey method will be checked unless Validity Check is cleared.