help canon postestimation dialog: predict estat
screeplot
also see: canon
-------------------------------------------------------------------------------
Title
[MV] canon postestimation -- Postestimation tools for canon
Description
The following postestimation commands are of special interest after
canon:
command description
-------------------------------------------------------------------------
estat correlations show correlation matrices
estat loadings show loading matrices
estat rotate rotate raw coefficients, standard coefficients, or
loading matrices
estat rotatecompare compare rotated and unrotated coefficients or
loadings
screeplot plot canonical correlations
-------------------------------------------------------------------------
The following standard postestimation commands are also available:
command description
-------------------------------------------------------------------------
estat VCE and estimation sample summary
estimates cataloging estimation results
lincom point estimates, standard errors, testing, and inference for
linear combinations of coefficients
nlcom point estimates, standard errors, testing, and inference for
nonlinear combinations of coefficients
predict predictions, residuals, influence statistics, and other
diagnostic measures
predictnl point estimates, standard errors, testing, and inference for
generalized predictions
test Wald tests of simple and composite linear hypotheses
testnl Wald tests of nonlinear hypotheses
-------------------------------------------------------------------------
Special-interest postestimation commands
estat correlation displays the correlation matrices calculated by canon
for varlist1 and varlist2 and between the two lists.
estat loadings displays the canonical loadings computed by canon.
estat rotate performs orthogonal varimax rotation of the raw
coefficients, standard coefficients, or canonical loadings. Rotation is
calculated on the canonical loadings regardless of which coefficients or
loadings are actually rotated.
estat rotatecompare displays the rotated and unrotated coefficients or
loadings and the most recently rotated coefficients or loadings. This
command may be used only if estat rotate has been performed first.
Syntax for predict
predict [type] newvar [if] [in] , statistic* [correlation(#)]
statistic* description
-------------------------------------------------------------------------
Main
u calculate linear combination of varlist1
v calculate linear combination of varlist2
stdu calculate standard error of the linear combination of
varlist1
stdv calculate standard error of the linear combination of
varlist2
-------------------------------------------------------------------------
* There is no default statistic; you must specify one statistic from the
list.
These statistics are available both in and out of sample; type predict
... if e(sample) ... if wanted only for the estimation sample.
Menu
Statistics > Postestimation > Predictions, residuals, etc.
Options for predict
+------+
----+ Main +-------------------------------------------------------------
u and v calculate the linear combinations of varlist1 and varlist2,
respectively. For the first canonical correlation, u and v are the
linear combinations having maximal correlation. For the second
canonical correlation, specified in predict with the correlation(2)
option, u and v have maximal correlation subject to the constraints
that u is orthogonal to the u from the first canonical correlation,
and v is orthogonal to the v from the first canonical correlation.
The third and higher correlations are defined similarly. Canonical
correlations may be chosen either with the lc() option to canon or by
specifying the correlation() option to predict.
stdu and stdv calculate the standard errors of the respective linear
combinations.
correlation(#) specifies the canonical correlation for which the
requested statistic is to be computed. The default value for
correlation() is 1. If the lc() option to canon was used to
calculate a particular canonical correlation, then only this
canonical correlation is in the estimation results. You can obtain
estimates for it by either specifying correlation(1) or by omitting
the correlation() option.
Syntax for estat
Display the correlation matrices
estat correlation [, format(%fmt) ]
Display the canonical loadings
estat loadings [, format(%fmt) ]
Perform orthogonal varimax rotation
estat rotate [, rawcoefs stdcoefs loadings format(%fmt) ]
Display the Rotated and unrotated coefficients or loadings
estat rotatecompare [, format(%fmt) ]
Menu
Statistics > Postestimation > Reports and statistics
Options for estat
format(%fmt) specifies the display format for numbers in matrices; see
[D] format. format(%8.4f) is the default.
rawcoefs, an option for estat rotate, requests the rotation of raw
coefficients. It is the default.
stdcoefs, an option for estat rotate, requests the rotation of
standardized coefficients.
loadings, an option for estat rotate, requests the rotation of the
canonical loadings.
Examples
Setup
. sysuse auto
. canon (mpg price weight) (length trunk turn)
Show loading matrices
. estat loadings
Show correlation matrices
. estat corr
First and second linear combinations of first varlist
. predict u1, u corr(1)
. predict u2, u corr(2)
Rotate standardized coefficients
. estat rotate, stdcoefs
Saved results
estat correlations saves the following in r():
Matrices
r(corr_var1) correlations for varlist_1
r(corr_var2) correlations for varlist_2
r(corr_mixed) correlations between varlist_1 and varlist_2
estat loadings saves the following in r():
Matrices
r(canload11) canonical loadings for varlist_1
r(canload22) canonical loadings for varlist_2
r(canload21) correlations between varlist_2 and the canonical
variates for varlist_1
r(canload12) correlations between varlist_1 and the canonical
variates for varlist_2
estat rotate saves the following in r():
Macros
r(coefficients) coefficients rotated
r(class) rotation classification
r(criterion) rotation criterion
Matrices
r(AT) rotated coefficient matrix
r(T) rotation matrix
Also see
Manual: [MV] canon postestimation
Help: [MV] canon;
[MV] rotatemat, [MV] screeplot
