Statscraft 0.7.0, release v0.7.0

DISCRIMINANT

Linear discriminant analysis with direct or stepwise entry and classification of cases.

Syntax

DISCRIMINANT GROUPS=var(min,max) /VARIABLES=varlist [/SELECT=var(value)]
   [/ANALYSIS=varlist[(level)] ...] [/METHOD=DIRECT | WILKS | MAHAL | MAXMINF | MINRESID | RAO]
   [/TOLERANCE=0.001] [/MAXSTEPS=n] [/FIN=3.84] [/FOUT=2.71] [/PIN=p] [/POUT=p] [/VIN=0]
   [/FUNCTIONS=n1[,pct[,sig]]] [/PRIORS=EQUAL | SIZE | values] [/SAVE=[CLASS[=name]] [SCORES[=root]] [PROBS[=root]]]
   [/ANALYSIS=... another block] [/MISSING=EXCLUDE | INCLUDE] [/HISTORY=STEP | NONE]
   [/CLASSIFY=[NONMISSING | UNSELECTED | UNCLASSIFIED] [POOLED | SEPARATE] [MEANSUB]]
   [/STATISTICS=[MEAN] [STDDEV] [UNIVF] [COV] [CORR] [FPAIR] [BOXM] [GCOV] [TCOV] [RAW] [COEFF] [TABLE]
                [CROSSVALID] [ALL]] [/ROTATE=NONE] [/PLOT=...].
The analysis uses the cases with a whole group value in the range, no missing value on VARIABLES and,
with SELECT, the selected value; groups without cases are ignored. DIRECT enters the variables that
pass the tolerance; the stepwise methods enter the variable with the best criterion (smallest Wilks'
lambda, largest smallest Mahalanobis distance or F between groups, smallest sum of unexplained
variation, largest increase in Rao's V) among those passing FIN or PIN, higher inclusion levels
first (even levels together, level 0 never), and remove a level 1 variable below FOUT or above POUT.
Tables: Analysis Case Processing Summary, Group Statistics, Tests of Equality of Group Means (UNIVF),
Pooled Within-Groups Matrices (COV, CORR), Covariance Matrices (GCOV, TCOV), Log Determinants and Box's
M (BOXM), the stepwise tables (HISTORY), Pairwise Group Comparisons (FPAIR), Eigenvalues, Wilks'
Lambda, Standardized Canonical Discriminant Function Coefficients, Structure Matrix, Canonical
Discriminant Function Coefficients (RAW), Functions at Group Centroids, Classification Processing
Summary, Prior Probabilities for Groups, Classification Function Coefficients (COEFF) and
Classification Results (TABLE, CROSSVALID leaving each case out). SAVE adds the predicted group, the
function scores and the posterior probabilities, named DSC_m and DSCn_m unless given. Weighted when a
weight variable is set. PLOT is accepted with a warning that it is not produced yet. Not supported
yet: ROTATE=COEFF and STRUCTURE, MATRIX, OUTFILE.

Example

DISCRIMINANT GROUPS=segment(1,3) /VARIABLES=q1 TO q6 /METHOD=WILKS /PRIORS=SIZE /STATISTICS=MEAN STDDEV UNIVF BOXM TABLE /SAVE=CLASS SCORES PROBS.

statscraft help DISCRIMINANT prints the same text in the terminal. All commands