help nestreg dialog: nestreg
-------------------------------------------------------------------------------
Title
[R] nestreg -- Nested model statistics
Syntax
Standard estimation command syntax
nestreg [, options ] : command_name depvar (varlist) [(varlist) ...]
[if] [in] [weight] [, command_options]
Survey estimation command syntax
nestreg [, options ] : svy [vcetype] [, svy_options ] : command_name
depvar (varlist) [(varlist) ...] [if] [in] [,
command_options]
options description
-------------------------------------------------------------------------
Reporting
waldtable report Wald test results; the default
lrtable report likelihood-ratio test results
quietly suppress any output from command_name
store(stub) store nested estimation results in _est_stub#
-------------------------------------------------------------------------
by is allowed; see prefix.
Weights are allowed if command_name allows them; see weight.
A varlist in parentheses indicates that this list of variables is to be
considered as a block. Each variable in a varlist not bound in
parentheses will be treated as its own block.
All postestimation commands behave as they would after command_name
without the nestreg prefix; see the postestimation manual entry for
command_name.
Menu
Statistics > Other > Nested model statistics
Description
nestreg fits nested models by sequentially adding blocks of variables and
then reports comparison tests between the nested models.
Options
+-----------+
----+ Reporting +--------------------------------------------------------
waldtable specifies that the table of Wald test results be reported.
waldtable is the default.
lrtable specifies that the table of likelihood-ratio tests be reported.
This option is not allowed if pweights, the vce(robust) option, or
the vce(cluster clustvar) option is specified. lrtable is also not
allowed with the svy prefix.
quietly suppresses the display of any output from command_name.
store(stub) specifies that each model fit by nestreg be stored under the
name _est_stub#, where # is the nesting order from first to last.
Remarks
Remarks are presented under the following headings:
Estimation commands
Wald tests
Likelihood-ratio tests
Programming for nestreg
Estimation commands
nestreg removes collinear predictors and observations with missing values
from the estimation sample before calling command_name.
The following Stata commands are supported by nestreg:
clogit, cloglog, glm, intreg, logistic, logit, nbreg, ologit,
oprobit, poisson, probit, qreg, regress, scobit, stcox, stcrreg,
streg, tobit
You do not supply a depvar for streg, stcox, or stcrreg; otherwise,
depvar is required. You must supply two depvars for intreg.
Wald tests
Use nestreg to test the significance of blocks of predictors, building
the regression model one block at a time. Using the data from the first
example of [R] test, we wish to test the significance of the following
predictors of birth rate: medage, medagesq, and region (already
partitioned into four indicator variables, reg1, reg2, reg3, and reg4).
. webuse census4
. nestreg: regress brate (medage) (medagesq) (reg2-reg4)
This single call to nestreg ran regress three times, adding a block of
predictors to the model for each run as in
. regress brate medage
. regress brate medage medagesq
. regress brate medage medagesq reg2-reg4
nestreg collected the F statistic for the corresponding block of
predictors and the model R-squared statistic from each model fit.
The F statistic for the first block, 164.72, is for a test of the joint
significance of the first block of variables; it is simply the F
statistic from the regression of brate on medage. The F statistic for
the second block, 35.25, is for a test of the joint significance of the
second block of variables in a regression of both the first and second
blocks of variables. In our example it is an F test of medagesq in the
regression of brate on medage and medagesq. Similarly, the third block's
$F$ statistic of 8.85 corresponds to a joint test of reg2, reg3, and reg4
in the final regression.
Likelihood-ratio tests
The nestreg command provides a simple syntax for performing
likelihood-ratio tests for nested model specifications; also see [R]
lrtest. Using the data from the first example of [R] lrtest, we wish to
jointly test the significance of the following predictors of low
birthweight: age, lwt, ptl, and ht.
. webuse lbw
. xi: nestreg, lr: logistic low (i.race smoke ui) (age lwt ptl ht)
The estimation results from the full model are left in e(), so we can
later use estat and other postestimation commands.
. estat gof
Programming for nestreg
If you want your user-written command (command_name) to work with
nestreg, it must follow standard Stata syntax and allow the if qualifier.
Furthermore, command_name must have sw or swml as a program property; see
> [P] program properties. If command_name has swml as a property,
command_name must save the log-likelihood value in e(ll) and the model
degrees of freedom in e(df_m).
Saved results
nestreg saves the following in r():
Matrices:
r(wald) matrix corresponding to the Wald table
r(lr) matrix corresponding to the likelihood-ratio table
Also see
Manual: [R] nestreg
Help: [P] program properties
