We have developed computer code that performs large sample analytic
predictor sort power calculations and checks some of these estimates
for small samples using simulations. The large sample results and
the small sample simulation programs are based upon research reported in
Verrill (1993), "Predictor Sort Sampling, Tight *T*'s, and the
Analysis of Covariance," *The Journal of the
American Statistical Association*, **88**, No. 421, 119-124
(pdf)
and
Verrill and
Kretschmann (2015), "A Reminder about
Potentially Serious Problems with a Type of Blocked ANOVA
Analysis," USDA Forest Products Laboratory draft research paper
(pdf).

The programs
can be run over the Web at
http://www1.fpl.fs.fed.us/sampsize.html.
The material below provides
links to FORTRAN versions of this computer code.

Disclaimer of Warranties

The FORTRAN sssimp source code (power calculations in the simple 1 factor, 2 levels case) comes in two files:

- sssimp.f --- A main routine written by Steve Verrill.
- quadint.f --- A quadratic interpolation subroutine written by Steve Verrill.

The sssimp program also depends upon DCDFLIB.
DCDFLIB is a
public domain library of "routines for cumulative distribution
functions, their inverses, and their parameters." It was produced
by Barry Brown, James Lovato, and
Kathy Russell of the Department of Biomathematics,
M.D. Anderson Cancer Center, The University of Texas.
DCDFLIB can be found
at
https://biostatistics.mdanderson.org/SoftwareDownload/SingleSoftware.aspx?Software_id=21
.

The FORTRAN sscomp.2 source code for power calculations involving a comparison of 2 pre-specified levels of a factor in the complex case (possibly more than two factors, possibly more than two levels per factor) comes in one file:

- sscomp.2.f --- A main routine written by Steve Verrill.

The sscomp.2 program also depends upon the DCDFLIB library described
above.

The FORTRAN simcomp.2 source code for simulation estimates of power for a comparison of 2 pre-specified levels of a factor in the complex case (possibly more than two factors, possibly more than two levels per factor) comes in four files:

- simcomp.2.f --- A main routine written by Steve Verrill.
- rnor.f --- rnor is a routine that "generates quasi normal random numbers, with mean zero and unit standard deviation." It was written by David Kahaner and George Marsaglia.
- rsort.f --- A routine that sorts real numbers. Written by Steve Verrill.
- slat.f --- Routines from the public domain SLATEC package of numerical analysis routines.

The simcomp.2 program also depends upon the DCDFLIB library described
above, and on the public domain LAPACK package of linear algebra
numerical analysis routines. This library can be found at
http://www.netlib.org/lapack/.

The FORTRAN sscomp.mult source code for power calculations involving a comparison of all levels of a factor in the complex case (possibly more than two factors, possibly more than two levels per factor) comes in one file:

- sscomp.mult.f --- A main routine written by Steve Verrill.

The sscomp.mult program also depends upon the DCDFLIB library described above.

The FORTRAN simcomp.mult source code for simulation estimates of power for a comparison of all levels of a factor in the complex case (possibly more than two factors, possibly more than two levels per factor) comes in four files:

- simcomp.mult.f --- A main routine written by Steve Verrill.
- rnor.f --- rnor is a routine that "generates quasi normal random numbers, with mean zero and unit standard deviation." It was written by David Kahaner and George Marsaglia.
- rsort.f --- A routine that sorts real numbers. Written by Steve Verrill.
- slat.f --- Routines from the public domain SLATEC package of numerical analysis routines.

The simcomp.mult program also depends upon the DCDFLIB library described
above, and on the public domain LAPACK package of linear algebra
numerical analysis routines. This library can be found at
http://www.netlib.org/lapack/.

Last modified on 5/16/15.

As of last midnight, this page had been accessed times.