STAT <data>,L
computes certain basic statistics and summaries for active variables in
<data> and forms a frequency distribution for each of them according to
an automatic classification.
L (optional) is the first line for the results.
The IND, CASES and SELECT specifications may limit the observations and the
missing values are recorded for each variable separately.
The main application of STAT is give an crude, overall account on
the behaviour of active variables. On the basis of the STAT results
decisions about more refined analysis can be done.
The basic statistics given by STAT include, for example, min, max,
mean, stddev, skewness, kurtosis, first order autocorrelation,
median, quartiles and entropy.
The selection of statistics depends, however, on the scale type of the
variable. Certain statistics (like autocorrelation) are also omitted,
if no information seems to be gained (autocorrelation<2/sqrt(N)).
Sums, sums of squares, etc. are computed by giving a specification
SUMS=<integer> where <integer>=1,2,3, or 4.
Thus by SUMS=1 plain sums are computed and displayed as sum1=<value>.
Special means of the form ((X1^k+X2^k+...+XN^k)/N)^(1/k) will be
computed for positive variables X and for various values of the
exponent k=k1,k2,... by entering a specification MEANS=k1,k2,...
Thus Quadratic means are obtained by k=2 or k=Q, Harmonic means by
k=-1 or k=H, Arithmetic means by k=1 or k=A, and Geometric means by
k=0 or k=G. For example, all these means plus a power mean with the
exponent k=2.5 are obtained by MEANS=A,G,H,Q,2.5
More fractiles are computed (from automatically grouped data) by
the specification FRACTILES=p1,p2,... where 0<p<1, for p=p1,p2,...
Order statistics are not given if (due to an outlier etc.) most of
the data values are accumulated into a single class.
Confidence intervals for means (under normality assumption) are obtained
by specification CONFMEAN, (Example: CONFMEAN=0.95).
In frequency distributions, maximum number of classes is given by the
CLASSMAX=<#_of_classes> specification (default CLASSMAX=30)
When making frequency distributions, STAT first tries to keep record
on each variable for each distinct value up to the limit CLASSMAX.
If it is not possible, a classification with equal class widths is
selected on the basis of values found so far. This classification
will be further redefined, by doubling the class widths, if values
outside the current range is found.
Values of string variables are considered up to # of characters given by
the MAXSTRING specification (default MAXSTRING=8).
The printout of frequency distributions is omitted if RESULTS<=30.
For grouped distributions with user-defined classes see HISTO and TAB.
A = Simple statistics (means, stddevs, frequency distributions)
B = Basic statistics
S = general information on statistical operations
F = STAT command for giving information about the status of the data file
M = Results of STAT as a matrix file STAT.M
Help System (web edition)