6.0 PROBIT Command
The PROBIT command allows estimation of dichotomous (0-1 left hand
variable) probit models with up to 97 variables on the right. A related
procedure is MPROBIT. Probit models can also be estimated from the
MATRIX command and nonlinear PROBIT models camn be estimated using
the MARSPLINE procedure. This latter capability is ahown in the
file probit_2 in example.mac whichj should be consulted for further
detail.
The general form of the PROBIT command is:
B34SEXEC PROBIT options parameters$
MODEL Yvar = Xvar1 Xvar2 ... $
MEANS Xvar1(value1,value2, ...) Xvar2(value1,value2, ...) $
B34SEEND$
The MODEL sentence is required.
PROBIT options:
ISECD - Specified if want output of second derivatives.
IITLK - Specified if want output of log likelihood after
each iteration.
IIEST - Specified if desire estimates after each iteration.
NOINT - Estimate model without an intercept.
SPUNCHRES - Places residuals on unit 44 in SCA FSAVE
format to a file with name RESIDUAL. Series listed
are OBSNUM, Y, YHAT, RESIDUAL, DENSITY. The estimated
model is listed in the file as comments. File is not
rewound prior to saving. SCAINPUT must be used to
rename this file if SPUNCHRES is used for subsequent
PROBIT commands. If this is not done, only the
first file will be read.
PROBIT parameters:
NEQ=n1 - Sets equation number. Default = 1. Max = 99.
NADJ=n2 - Sets number of points for user supplied values of right
hand variables. If this option is supplied, n2 Z scores
and their associated probabilities are calculated. If
NADJ is supplied, the MEANS sentence must be supplied
and n2 values given for any specified variable listed.
The maximum value for NADJ = 20.
NSTRT=n3 Sets beginning observation for output of calculated and
actual dependent variable density. Default = 1.
NSTOP=n4 Sets ending observation for output of calculated and
actual dependent variable. Default = 5.
TOLA=value Sets convergence tolerence. Default = .00001. A Maximum
of 10 digits can be supplied.
MODEL sentence.
MODEL Y = X1 X2 X3 $
Where Y is the left hand variable and X1 X2 X3 are right hand variables.
Y, X1, X2 X3 must have been passed to B34S. The maximum number of right
hand variables = 48.
MEANS sentence (only needed if NADJ parameter is supplied).
MEANS Var1(value1,value2,..,valuen2),Var2(value1,value2,...,valuen2) $
Note: Not all variables must be mentioned in the MEANS sentence but if
any variable is mentioned, all NADJ values should be given. For each
value supplied, a maximum of 8 digits (including required decimal point)
is allowed. For further detail on the MEANS sentence, see NADJ
parameter.
Sample setup of PROBIT.
B34SEXEC PROBIT $
MODEL Y = X Q P $
B34SEEND$
B34SEXEC PROBIT NADJ=3 $
MODEL Y = X P Q$
MEANS X(.50,.88,1.03) P(1.0,2.0,1.0) $
B34SEEND$
Note: In the second example 3 sets of Z scores and associated
probabilities are calculated. In calculating the Z scores, sample mean
values are used for Q and the constant, the X values used are .50, .88
and 1.03 and the P values used are 1.0, 2.0 and 1.0. For example
Z(1) = B(1) + B(2) * .50 + B(3) * 1.0 + B(4) * (mean Q)
Z(2) = B(1) + B(2) * .88 + B(3) * 2.0 + B(4) * (mean Q)
Z(3) = B(1) + B(2) *1.03 + B(3) * 1.0 + B(4) * (mean Q)
where B(i), i=1,4 are the estimated coefficients.
The Z(i) values can be converted into probabilities with the SAS or B34S
function PROBNORM. For further information on how to use PROBNORM, see
the B34S DATA command.