.-
help for ^mlogpred^                                        (statalist: 17sep1998)
.-

Predicted probabilities after ^mlogit^ and ^svymlog^
-------------------------------------------------

    ^mlogpred^ varname1 [varname2] [^if^ exp] [^in^ range] [^,^ { ^p^rob | ^se^ | ^ci^ }
                                 { ^e^quation^(^#^)^ | ^o^utcome^(^#^)^ } ^l^evel^(^#^)^ ]


^mlogpred^ works after ^mlogit^ and ^svymlog^.


Description
------------

^mlogpred^ computes predicted probabilities, standard errors of predicted
probabilities, and confidence intervals for predicted probabilities after
^mlogit^ and ^svymlog^.


Options
--------

^prob^ requests that predicted probabilities be placed in varname1.  If none of
    ^prob^, ^se^, or ^ci^ is specified, then ^prob^ is the default.

^se^ requests that standard errors of predicted probabilities be placed in
    varname1.

^ci^ requests that confidence intervals of predicted probabilities be computed.
    The lower bound of the confidence interval is placed in varname1, and the
    upper bound is placed in varname2.

^equation(^#^)^ (synonym ^outcome(^#^)^) specifies the outcome for which the predicted
    probabilities, standard errors, or confidence intervals are to be computed.
    The # is either (1) an equation number, e.g., #1, #2,..., (2) the numeric
    value of the outcome, or (3) the value label of the outcome.

    For both ^mlogit^ and ^svymlog^ equation numbers (#1, #2,...) can always be
    used to specify outcomes.

    For ^mlogit^, numeric values of the outcome can always be used.  If ^mlogit^
    uses value labels to label equations, then value labels can also be used
    to specify outcomes.

    For ^svymlog^, numeric values can only be used if equations on the ^svymlog^
    output are labeled with numeric values.  If ^svymlog^ uses value labels to
    label the equations, then value labels must be used rather than numeric
    values.

^level(^#^)^ specifies the confidence level (i.e., nominal coverage rate), in
    percent, for confidence intervals; see help @level@.


Technical note
---------------

Confidence intervals for predicted probabilities are constructed using a logit
transform so that their endpoints always lie between 0 and 1.


Examples
---------

	. ^mlogit y x1 x2 x3^

	. ^mlogpred phat, equation(#1)^
	. ^mlogpred phat, outcome(1)^
	. ^mlogpred phat, prob out(Good)^

	. ^mlogpred sehat, se eq(#1)^

	. ^mlogpred lb ub, ci eq(#1)^

	. ^svymlog choice age income^

	. ^mlogpred p1, equation(#1)^
	. ^mlogpred red, outcome(red car)^

	. ^mlogpred error, se outcome(blue car)^

	. ^mlogpred ci1 ci2, ci outcome(purple truck)^


Author
-------

	Bill Sribney
	Stata Corporation
	702 University Drive East
	College Station, TX 77840
	email: wsribney@@stata.com


Also see
---------

    STB:  ^STB-40 svy6 (svymlog)^
 Manual:  ^[R] mlogit^
          ^[R] predict^
	  ^[R] svy^
On-line:  help for @mlogit@, @svymlog@ (if installed); @predict@, @svy@