home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Between Heaven & Hell 2
/
BetweenHeavenHell.cdr
/
500
/
473
/
multi.arc
/
PC-MULTI.DOC
< prev
next >
Wrap
Text File
|
1985-10-16
|
9KB
|
281 lines
PC-MULTI
Multiple Comparisons
Version 1.0
October 15, 1985
Gerard E. Dallal
USDA Human Nutrition Research Center on Aging
at Tufts University
711 Washington Street
Boston, MA 02111
and
Tufts University School of Nutrition
132 Curtis Street
Medford, MA 02155
PC-MULTI constructs simultaneous confidence intervals for
pairwise mean differences using Tukey's honest significant
differences (Studentized range statistic).
NOTICE
Documentation and original code copyright 1985 by Gerard E.
Dallal. Reproduction of material for non-commercial purposes
is permitted, without charge, provided that suitable
reference is made to PC-MULTI and its author.
Neither PC-MULTI nor its documentation should be modified in
any way without permission from the author, except for those
changes that are essential to move PC-MULTI to another
computer.
Please acknowledge PC-SIZE in any manuscript that uses its
calculations.
DESCRIPTION
PC-MULTI constructs simultaneous confidence intervals for
pairwise mean differences using Tukey's honest significant
differences (Studentized range statistic). See Miller (1981,
sec.2.1) or Winer (1971, sec.3.8).
PAGE 2
PC-MULTI produces two sections of output:
Confidence intervals based on Studentized
range statistic
For each pairwise difference, PC-MULTI reports the sample
mean difference and the upper and lower confidence limits for
the population mean difference (sample mean difference plus
and minus an honest significant difference). When sample
sizes are unequal, PC-MULTI uses the suggestion of
Kramer(1956):
(sample mean difference) +-
Q * SQRT(MSE * (1/N1 + 1/N2) / 2),
where Q is the critical value based on the Studentized range
statistic, MSE is the mean square error, and N1 and N2 are
the sample sizes. See Dunnett (1980) for a simulation study
of the behavior of this procedure.
A diagram to the right of the statistics gives a graphical
display of the differences among the groups. The vertical
line in the diagram indicates the location of 0 on the
horizontal axis. The confidence intervals terminate in
parentheses (when the lower limit is greater than zero or the
upper limit is less than zero) or square brackets (when the
upper limit is at least zero or the lower limit is no greater
than zero). This device allows for easy identification of
non-significant confidence intervals when, due to rounding,
the end of a confidence interval falls in the same location
as the zero line. Think of the square bracket as "grabbing"
0, e.g.,
|
(----- significant (does not include 0)
----] not significant (includes 0)
|
Using harmonic mean...
For each group PC-PITMAN reports its mean plus and minus half
an honest significant difference, computing the HSD's using
the harmonic mean of the sample sizes (Winer,1971,p.216).
The diagram to the right of the statistics gives a graphical
display of this information. Two groups whose intervals
PC-MULTI G.E. Dallal
PAGE 3
overlap are not significantly different. If their intervals
do not overlap, they ARE significantly different.
The appeal of this diagram is that groups whose intervals
overlap are not significantly different while intervals whose
intervals do not overlap ARE significantly different.
INSTALLATION
PC-SIZE is written for the IBM-PC. Installation on a new
computer may entail modifying the following statements:
The first DATA statement:
IIN -- input unit number (screen)
IOUT -- output unit number (screen)
IWOUT -- save file unit number
MAXGRP -- largest number of sample means that
can be handled
The OPEN statement for the save file just before statement
10.
If MAXGRP is modified, the dimensions of DATA, GNAME, XNOBS,
CIM, CIL, CIU, CILH, CIUH should be adjusted accordingly in
the main program.
OPERATION
Data may be entered through the keyboard or from an external
file.
Data entry through the keyboard
PC-MULTI provides all of the necessary prompts. Each group
name may be up to six characters in length and must be
enclosed in single quotes.
If a zero is entered in response to the degrees of freedom
for error prompt, PC-MULTI will calculate the error degrees
PC-MULTI G.E. Dallal
PAGE 4
of freedom as the total number of observations minus the
number of groups (which is the case for a standard one-way
layout).
Format of external data files
Record 1: The first column must contain an 'E' or 'e' (no
quotes). The error variance must appear within the next
eighteen columns.
Record 2: The first column must contain an 'D' or 'd' (no
quotes). The degrees of freedom for the error term must
appear within the next eighteen columns. If the value is
zero or blank PC-MULTI will calculate the error degrees
of freedom as the total number of observations minus the
number of groups (which is the case for a standard one-
way layout).
Records 3 thru end: Each record must contain three entries--
sample mean, sample size, and group name--separated by
spaces or commas. Group names must be enclosed in single
quotes.
ALGORITHMS
PC-SIZE makes use of the following published routines:
Lund, R.E. and J.R. Lund (1983). Algorithm AS 190.
Probabilities and upper quantiles for the Studentized
range. Appl. Statist.,32,204-210. However, the function
QTRANG (Algorithm AS190.1) is defective and was replaced
by a function of the author's construction.
Hill, I.D. (1973). Algorithm AS 66. The normal integral.
Appl. Statist.,22,424-427.
Odeh, R.E. and J.O. Evans (1974). Algorithm AS 70. The
percentage points of the normal distribution. Appl.
Statist.,23,96-97.
PC-MULTI G.E. Dallal
PAGE 5
REFERENCES
Dunnett, Charles W. (1980). Pairwise multiple comparisons in
the homogeneous variance, unequal sample size case.
J. Amer. Statist. Assn.,75,789-795.
Kramer, C.Y. (1956). Extension of multiple range tests to
group means with unequal numbers of replications.
Biometrics,12,307-310.
Miller, Rupert G., Jr. (1981). Simultaneous Statistical
Inference, 2-nd ed. New York: Springer-Verlag, Inc.
Winer, B.J. (1971). Statistical Principles in Experimental
Design, 2-nd ed. New York: McGraw-Hill Book Company, Inc.
PC-MULTI G.E. Dallal