BOX

The BOX command does Box-Cox regressions.

In general, the format is:

  BOX depvar indeps / options


The available options are:

Option Description
ACCUR Iterates to an accuracy of .001 instead of .01 for lambda.
ALL Extends the Box-Cox model so that all variables receive the same power transformation.
AUTO Will simultaneously estimate lambda and the first-order autocorrelation parameter rho.
With this option the DROP, NMISS= and GAP= options as described for the AUTO command can be used.
DN Uses N instead of N-K as the divisor for SIGMA**2.
DUMP Gives some intermediate output on the iterations.
FULL Combined Box-Cox and Box-Tidwell estimation is done and all variables have different lambdas.
NOCONSTANT Forces the regression through the origin.
TIDWELL Box-Tidwell regression is done.
UT Untransforms the observed and predicted dependent variables.
COEF= Saves the coefficients, the lambda for each independent variable and the rho if there is autocorrelation in the matrix specified.
LAMBDA= Specifies a fixed value for lambda.
LAMS= Used to do a manual grid search of lambda. The LAME= starting value (LAMS=), the ending value (LAME=) and LAMI= an increment (LAMI=) of lambda should be specified.
RHO= Specifies the rho to be used when the AUTO option is specified.

Other available options are:
ANOVA, GF, LIST, MAX, PCOR, PCOV, RESTRICT, RSTAT, BEG=, END=, COV=, PREDICT= and RESID= as defined for OLS

The available temporary variables on the BOX command are:
$ANF, $DF, $DW, $ERR, $K, $LLF, $N, $R2, $R2OP, $RAW, $RHO, $SIG2, $SSE, $SSR, $SST, $ZANF, $ZDF, $ZSSR, $ZSST.