Statscraft 0.7.0, release v0.7.0

REGRESSION

Linear regression with forced entry or removal and stepwise selection, residuals and diagnostics.

Syntax

REGRESSION [/VARIABLES=varlist | (COLLECT)] [/DESCRIPTIVES=[DEFAULTS MEAN STDDEV VARIANCE CORR SIG
           BADCORR COV XPROD N ALL NONE]] [/MISSING=LISTWISE | PAIRWISE | MEANSUBSTITUTION [INCLUDE]]
           [/REGWGT=var] [/STATISTICS=[DEFAULTS R ANOVA COEFF OUTS ZPP CHANGE CI[(95)] BCOV COLLIN TOL
           F SES XTX SELECTION LABEL ALL]] [/CRITERIA=[DEFAULTS PIN[(0.05)] | FIN[(3.84)] POUT[(0.10)] |
           FOUT[(2.71)] TOLERANCE[(0.0001)] MAXSTEPS(n) CIN[(95)]]] [/ORIGIN | NOORIGIN]
           /DEPENDENT=varlist /[METHOD=]ENTER [varlist] | REMOVE varlist | STEPWISE [varlist] |
           FORWARD [varlist] | BACKWARD [varlist] [/METHOD=...]
           [/RESIDUALS=[DEFAULTS DURBIN OUTLIERS(tempvars) ID(var) HISTOGRAM NORMPROB]]
           [/CASEWISE=[DEFAULTS OUTLIERS(3) | ALL PLOT(tempvar) DEPENDENT tempvars]]
           [/SAVE=tempvar[(name)] ... [FITS]] [/SCATTERPLOT ...] [/PARTIALPLOT ...].
tempvar: PRED RESID ZPRED ZRESID SRESID DRESID SDRESID ADJPRED SEPRED MAHAL COOK LEVER DFBETA SDBETA
DFFIT SDFIT COVRATIO MCIN ICIN. One equation per dependent; each METHOD works on the equation the
ones before it left, and each step is a model. ENTER takes the variables in order of tolerance;
STEPWISE removes the weakest variable while its F-to-remove fails POUT, then enters the strongest
that passes PIN; FORWARD only enters, BACKWARD only removes (entering its list first when the
equation is empty). A variable enters only when it and the variables in the equation keep the
TOLERANCE. Without keywords STATISTICS is R ANOVA COEFF OUTS and DESCRIPTIVES is MEAN STDDEV CORR.
The tables: Variables Entered/Removed, Model Summary (CHANGE, DURBIN, SELECTION add columns),
ANOVA, Coefficients (CI, ZPP, TOL, COLLIN, SES, F add columns), Excluded Variables, Coefficient
Correlations (BCOV), Collinearity Diagnostics (COLLIN), the swept correlation matrix (XTX),
Residuals Statistics (with RESIDUALS, CASEWISE or a plot), Casewise Diagnostics and Outlier
Statistics. LISTWISE (default) uses the cases valid on every variable; PAIRWISE builds the
correlations pair by pair, with the smallest pair count as N; MEANSUBSTITUTION puts the mean in
place of a missing value. SAVE adds variables named PRE_1, RES_1 and so on, or as given; DFBETA and
SDBETA save one per coefficient (name0, name1, ...), MCIN and ICIN a lower and an upper bound.
Weighted when a weight variable is set; REGWGT gives weighted least squares. HISTOGRAM, NORMPROB,
SCATTERPLOT, PARTIALPLOT and TEMPLATE are accepted with a warning that charts are not drawn yet.
LABEL changes nothing. Not supported yet: METHOD=TEST, SELECT, MATRIX, OUTFILE.

Example

REGRESSION /STATISTICS=COEFF OUTS R ANOVA CHANGE /DEPENDENT=spend /METHOD=ENTER age /METHOD=STEPWISE q1 TO q5 /SAVE=PRED ZRESID.

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