STEPPROBIT Procedure

@StepProbit does a backwards stepwise reduction of a probit model, dropping variables (one at a time) as long as the smallest t-statistic is less than a chosen threshold in absolute value.

@StepProbit( options ) depvar start end

# list of explanatory variables(in regression format)

Parameters

depvar	dependent variable
start, end	range for estimation. By default, the maximum range permitted by all variables involved.

Options

THRESHOLD=t-statistic required for a variable to stay in the model [1.5]

FORCE=number of regressors to include in all models [0]

By default, all variables listed on the supplementary card are treated identically. Any may be included or omitted from the final regression. To force regressors into every model, list those variables first on the supplementary card and use the option FORCE to indicate how many are to be forced in. If some of these variables are handled using lag fields, count the number of actual regressors.

[PRINT]/NOPRINT

TITLE="title for output" ["Stepwise Probit"]

Use NOPRINT to suppress output. Use TITLE to add to the output a description of what is being computed.

Variables Defined

These will be the variables defined by the final probit estimates, done with DDV.

Example

This applies @StepProbit to the model estimated in PROBIT.RPF, forcing the CONSTANT (regressor in the first position) into every model.

@stepprobit(force=1) yesvm

# constant public1_2 public3_4 public5 private years teacher loginc logproptax

Sample Output

The stepwise procedure has removed PUBLIC1_2, PUBLIC5, PRIVATE and YEARS as explanatory variables. Note that it would also have removed CONSTANT (which has a t-stat of just .8 in absolute value, which is less than the default threshold of 1.5), had it not been for the FORCE option.

Stepwise Probit - Estimation by Newton-Raphson

Convergence in 6 Iterations. Final criterion was 0.0000022 <= 0.0000100

Dependent Variable YESVM

Usable Observations 95

Degrees of Freedom 90

Log Likelihood -51.2126

Average Likelihood 0.5832849

Pseudo-R^2 0.2309487

Log Likelihood(Base) -62.5204

LR Test of Coefficients(4) 22.6158

Significance Level of LR 0.0001511

Variable Coeff Std Error T-Stat Signif

************************************************************************************

1. Constant -3.255070743 4.067369394 -0.80029 0.42354341

2. PUBLIC3_4 0.608227803 0.335213248 1.81445 0.06960842

3. TEACHER 1.657255062 0.913908847 1.81337 0.06977474

4. LOGINC 1.631594722 0.490594548 3.32575 0.00088181

5. LOGPROPTAX -1.860401428 0.664269011 -2.80067 0.00509959