find.ncp {SharedHT2} | R Documentation |
In order to provide a yardstick in designing simulation experiments this function calculates the non-centrality parameter for a simple F test of the null hypothesis of zero group means
find.ncp(e.I, e.II, nreps, d)
e.I |
type I error i.e. level of significance |
e.II |
type II error i.e. 1-power |
nreps |
# replicates |
d |
# groups |
An numeric vector of length 2, containing the calculated
effect-size
and the corresponding power of the standard F test
of the null hypothesis that all group means are zero at the alternative
that each group has mean effect-size
times the common within
group standard deviation.
theta |
The effect size |
power |
Corresponding power |
This is used to set up the true positives genes in the simulation routines by deciding which genes are ``true positives" in tests of the null that all groups means are zero. Each group is assigned a mean equal to $σ theta$ where $σ$ is the average common within group standard deviation and $theta$ is the effect size returned by a call to this function, for example $theta=2.0$ as in the example below. In calls to the simulation routine you only need to specify the per gene effect size as $σ$ is calculated internally in all cases. For example if you want the first 100 rows to be ``true-positives" in simulated micro-array data containing 15000 genes then in the simulation routines just specify ``effect.size=c(rep(es, 100), rep(0, 15000-100)" in the calls to simulation routines listed above. See the documentation for routines listed above in ``see also".
Grant Izmirlian izmirlian@nih.gov
SimNorm.IG
, SimMVN.IW
,
SimMVN.mxIW
find.ncp(0.005, 0.13179983, 3, 2)