qdunnett(p, k, df, nvec, control, two.sided=T)
Inf
is accepted
(though specifying
df=Inf
will result in the use of
df=1000
).
Missing values are not allowed.
TRUE
a value for two-sided comparisons is returned.
Suppose X1, X2, ... Xk are independent normal variables with means 0 and
variances proportional to the entries of
1/nvec
.
Let U be a random variable independent of the X's,
such that df*U^2 is Chisquare with df degrees of freedom.
The two-sided pivotal quantity is defined to be
D = max{1<=i<=k,i!=c: |Xi-Xc|/(U*sqrt(1/nvec(i)+1/nvec(c))) },
where c denotes the control's subscript.
The one-sided pivotal quantity is the above without absolute values.
The function obtains the critical point by numerical integration
and a secant method.
Hochberg, Y. and Tamhane, A. C.(1987).
Multiple Comparison Procedures.
Wiley, New York.
Hsu, Jason C. (1996).
Multiple Comparisons: Theory and Methods.
Chapman and Hall, London.
qdunnett(.99, 10, 20, rep(3, 10), 10, two.side=F) qdunnett(.99, 10, 26, c(3, 3, 3, 3, 3, 3, 3, 3, 3, 9), 10, two.side=F)