home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
SPACE 2
/
SPACE - Library 2 - Volume 1.iso
/
apps
/
705
/
stat1.doc
< prev
next >
Wrap
Text File
|
1992-03-01
|
5KB
|
109 lines
STATS 1 MENU
NORMALITY TESTS
These two tests check whether a variable is drawn from a normal
population. In both cases, you must select the variable to operate
on. In the case where you know the parameters, you will be asked
for the mean and standard deviation of the underlying population.
The program will not evaluate the statistic returned. Published
tables must be used.
DESCRIPTIVE STATISTICS
These tests provide mean, standard deviation, kurtosis, etc., for
the data set. If using grouped data, you will have to select a
grouping variable as well as the data variable.
CORRELATION
These provide simple correlation tests.
-Simple correlation is the correlation between two variables which
you choose.
-The Spearman rank correlation test compares the ranks of two sets
of variables rather than the actual numbers.
-The contingency coefficient test compares two variables on a
parametric basis. Data must be non-negative and scaled nominally.
-Kendal Concordance is used with three or more variables which are
in the form of ranks. No selection of variables is made. The
entire set of data in memory is used.
-The Kendal Tau test is similar to the Spearman rank test. It is
used for two variables in the form of ranks.
-The Point Biserial correlation test is used with two related
variables. One variable is at least intervally scaled, and the
other is a dichotomous variable. A dichotomous variable is one
which can have only two values 0 or 1, such as for male versus
female. You will be asked separately for the two variables.
-Lagged Auto correlation determines the correlation of a variable
with itself at an earlier time. The program will ask for the
variable to be examined and a variable into which to store the
results. The result is a series of values specifying the
correlation for a multitude of lag periods. The first value is
with no lag and has value 1.
-Lagged Multiple Correlation is similar to the above except that
two variables are examined. In this case, you are asked for two
variables. Order of selection is important. The lag period will
refer to the value of the second variable "K" periods earlier.
Thus, if Variable "A" is to be related to Variable "B" at earlier
periods, you should select Variable "A" first and "B" second.
-Partial correlation measures the correlation of two variables
with the effect of another group of variables removed. You select
a group of variables. The result is a matrix where the off diagonal
elements are partial correlation coefficients and the diagonal
elements are multiple correlation coefficients.
You need at least 3 variables.
ORDINAL TESTS
-Kolmogorov-Smirnov test checks a single variable to determine if
the values support the hypothesis that the differences between
them are chance.
-Mann Whitney "U" test requires two independent samples
(variables). The variables do not have to be the same size.
The test determines whether there is a difference in the rankings
between two groups. Small values are extreme for this test so the
comparison is "is the value less than the table value?"
-Wilcoxon test is similar to the Mann Whitney, except that it uses
related or paired variables. Like the Mann Whitney, small values
are extreme. Thus if the calculated value is less than the tabular
value you reject the null hypothesis that there is no difference
between samples.
-Kruskal Wallis test uses all data in the data set. There must be
at least three variables. The test is basically the three-or-more-
factor equivalent of Mann Whitney. The extreme values are high,
unlike the Mann Whitney.
-Friedman test uses all the data. This is the three-or-more-factor
equivalent of Wilcoxon. However, because of the formulation, the
statistic extreme values are high.
-Median test indicates whether the two samples appear to be drawn
from populations with the same median.
-Runs test is used for one variable. In addition, you must select
the test criteria from among zero, the mean, and the median. The
test determines whether the data appears to be randomly
distributed about the criterion.
-Sign test is used to test the probablility that a given variable
has a median of some test value. You must choose a variable and
then a test value for the median.
NOMINAL TESTS
-Chi Square 1 test uses two variables, with the first representing
the expected number of occurrences and the second the actual
number. The test determines whether the actual data is consistant
with the expected.
-Chi Square 2 test uses all of the data. The data is assumed to be
set up in a contingency matrix form. The null hypothesis is that
there is no relationship between the rows and columns.
-McNemar test uses all data. There must be two rows and two
columns. It is a test used to investigate changes in response in a
pre- and post-stimulous study. See the manual for data setup.
-Cochran test uses all data in the data set. There must be three
or more related variables with dichotomous data. In B/STAT
positives are treated as 1, negatives or 0 as 0. The test uses the
null hypothesis that there is no difference between variables.