DataMuseum.dk

top - download

    Length: 2355 (0x933)
    Types: TextFile
    Names: »oneway.1«

└─⟦a0efdde77⟧ Bits:30001252 EUUGD11 Tape, 1987 Spring Conference Helsinki
    └─ ⟦this⟧ »EUUGD11/stat-5.3/eu/stat/man/oneway.1« 

TextFile

.TH ONEWAY 1 "December 24, 1986" "\(co 1985 Gary Perlman" "|STAT" "UNIX User's Manual"
.SH NAME
oneway \- one way analysis of variance / t-test
.SH SYNOPSIS
.B oneway
[-ptu] [-w plotwidth] [-s splitter] [names]
.SH DESCRIPTION
.I oneway
does a one-way analysis of variance on data from different groups.
These groups can have different cell sizes.
When cell sizes are equal across groups,
the unweighted means analysis and the weighted means solutions
produce identical results.
When cell sizes are unequal,
the preferred analysis depends on the reasons for unequal cell sizes.
A text should be consulted to decide which test is most appropriate.
.PP
The input format is simple:
each group's data is separated from other groups' by including
a special value, called the splitter.
.SH OPTIONS
.de OP
.TP
.B -\\$1 \\$2
..
.PP
Group names can be supplied on the command line.
Otherwise, names like ``Group-1'' are used.
.OP p
Print error bar plots of each group.
The wide line of dashes is two standard deviations on each side of the mean
of each group.
The narrower double line is one standard deviation on each side.
Single standard errors are marked with parentheses.
The mean of each group is marked with a pound sign.
The minimum and maximum values are delimited by angle brackets.
.OP s splitter
This value is used to tell when one group's data ends and another's begins.
.OP t
Request a between group's t-test format for the significance test.
This option is ignored if there are not two groups.
.OP u
Request unweighted means solution rather than the default unweighted.
.OP w plotwidth
Set the width of the error bar plot to a value other than the default 60.
.SH EXAMPLE
.P
Suppose you had two groups of data in two files, called ``long'' and ``short.''
To compare these, you would do something like:
.ce
echo 999 | cat long - short | oneway -s 999 long short
.SH ALGORITHMS
The weighted means analysis algorithm follows Guilford & Fruchter's
.I "Fundamental Statistics for Psychology and Education (5th Edition)."
The unweighted means solution follows Keppel's
.I "Design and Analysis: A Researcher's Handbook."
.SH LIMITS
Use the -L option to determine the program limits.
.SH "MISSING VALUES
Missing data values (NA) are counted but not included in the analysis.
.SH "SEE ALSO
rankind(1) for a non-parametric rank order version of the analysis.