Stata 6 printed manual errata

Errata are available for the following manuals:

Reference Manual, Volume 1

Reference Manual, Volume 2

Reference Manual, Volume 3

Reference Manual, Volume 4

User’s Guide

Reference Manual, Volume 1

(all) [R] arch, p. 76, Syntax
In the syntax diagram, an asterisk is required with the use of a "stub" in the score() option.

... score(newvarlist | stub) ... ... score(newvarlist | stub*) ...
(all) [R] arch, p. 81, Syntax for predict
Add yresiduals to the available statistics list for predict. Insert the following statement.

yresiduals   residuals or predicted innovations in y—the undifferenced series
(all) [R] arch, p. 81, Options, saarch(numlist) option, second paragraph
saarch() may not ... with tarch(), narch(), ... saarch() may not ... with narch(), ...
(all) [R] arch, p. 89, Options for predict. Add the following paragraph describing the yresiduals option.
yresiduals calculates the residuals in terms of depvar, even if the model was specified in terms of, say, D.depvar. As with residuals, the yresiduals are computed from the model including any ARMA component. If option structural is specified, any ARMA component is ignored and yresiduals are the residuals from the structural equation; see structural below.
(all) [R] arima, p. 111, Syntax, Full syntax
In the syntax diagram an asterisk is required with the use of a “stub” in the score() option.

... score(newvarlist | stub) ... ... score(newvarlist | stub*) ...
(all) [R] arima, p. 111, Syntax for predict
Add yresiduals to the available statistics list for predict. Insert the following statement.

yresiduals   residuals or predicted innovations in y—the undifferenced series
(all) [R] arima, p. 116, Options for predict
Add the following paragraph describing the yresiduals option.

yresiduals calculates the residuals in terms of depvar, even if the model was specified in terms of, say, D.depvar. As with residuals, the yresiduals are computed from the model including any ARMA component. If option structural is specified, any ARMA component is ignored and yresiduals are the residuals from the structural equation; see structural below.
(1,2) [R] arima, p. 125, Methods and Formulas, eq. (12)
Bad Eq. 12 Good Eq. 12
(1,2) [R] arima, p. 125, Methods and Formulas, eq. (13)
Bad Eq. 13 Good Eq. 13
(3) [R] arima, p. 125, Methods and Formulas, eq. (13)
Bad Eq. 13 Good Eq. 13
(all) [R] biprobit, p. 132, Syntax
A comma is missing in the description of the syntax for equation1 and equation2.

... [ varlist ] [ offset(varname) ... ... [ varlist ] [, offset(varname) ...
(all) [R] biprobit, p. 133, Description
biprobit estimates maximum-likelihood probit models with sample selection. biprobit estimates maximum-likelihood two-equation probit models—either a bivariate probit or a seemingly unrelated probit (limited to two equations).
(all) [R] ci, p. 194, Syntax
ci varlist [weight] ... ci [varlist] [weight] ...
(all) [R] confirm, p. 243, Syntax
Keyword new cannot be abbreviated.

confirm [new] file filename
confirm [new | ... ] variable [varlist]
confirm [new] file filename
confirm [new | ... ] variable [varlist]
(1) [R] cox, p. 264, Syntax, second sentence concerning weights below the syntax diagram
... use the exactm or the exactp options ... ... use the exactp or efron options ...
(all) [R] epitab, p. 404, Saved Results, ir and iri saved results list
r(p)   two-sided p-value r(p)   one-sided p-value
(all) [R] glogit, p. 537, Syntax, blogit and bprobit syntax diagrams
Both blogit and bprobit allow the in range syntax. In both syntax diagrams insert the optional in qualifier.

... [if exp] [, ... ... [if exp] [in range] [, ...

Reference Manual, Volume 2

(1) [R] heckman, p. 27, Methods and Formulas, middle of page
are obtained. From these estimates the Mills' ratio m_j for each observation j is computed as are obtained. From these estimates the nonselection hazard, what Heckman (1979) referred to as the inverse of the Mills' ratio, m_j for each observation j is computed as
(all) [R] heckman, p. 27, Methods and Formulas, bottom of page, definition of sigma-hat squared
Replace the Beta_m with Beta_m^2 in the formula defining sigma-hat squared.
(all) [R] kdensity, p. 150, Methods and Formulas, formula for the Parzen kernel
Replace the formula for the Parzen kernel with

Parzen K[z] = { 4/3 - 8z2 + 8|z|3
8(1 -|z|)3/3
0
if |z| <= 1/2
if 1/2 < |z| <= 1
otherwise
(1) [R] label, p. 163, Description
label define ... associations of nonnegative integers and text ... label define ... associations of integers and text ...
(all) [R] limits, pp. 175 and 178, Remarks
Some of the limits were incorrectly listed.

Number of dyadic operators in an expression 66 66
Number of numeric literals in an expression 50 50
Number of conditions in an if statement 30 30
reg3, sureg, and other system estimators, Number of equations 50 50
Number of dyadic operators in an expression 66 200
Number of numeric literals in an expression 50 150
Number of conditions in an if statement 30 100
reg3, sureg, and other system estimators, Number of equations 40 800
(1) [R] logit, p. 231, Example
This page incorrectly reprints the contents of page 230. Click here to view the correct content for page 231.
(all) [R] macro, p. 268, Syntax
Add the following extended macro functions.

pwd
dirsep
tsnorm op[.varname] [, varname]

The following paragraphs may be added to describe these extended macro functions.

pwd returns the current working directory using the directory separator that is natural for your computer ("/", ":", or "\") and without an ending separator.

dirsep returns the directory separator that will work for your computer (either "/" or ":" but never "\" since Stata understands "/" which avoids potential macro substitution problems).

tsnorm returns a raw operator list in normalized form. The varname option indicates that the argument contains an operator and a varname (rawop.varname) and that both are to be returned (op.varname).
(1) [R] matrix, p. 321, Syntax, list of internal_Stata_matrix_names
b _b
V VCE
(1) [R] ml, p. 388, Options for use with mlmatsum, level(#) option
Option level(#) should not appear since mlmatsum does not take this option.
(all) [R] nbreg, p. 423, Syntax
Add the following statement.

nbreg may be used with sw to perform stepwise estimation; see [R] sw.
(all) [R] net, p. 442, Remarks, Example 1, xyz.pkg
Replace the leading "p" on the last three lines of xyz.pkg with "f".

p xyz.ado
p xyz.hlp
p sample.dta
f xyz.ado
f xyz.hlp
f sample.dta
(all) [R] net, p. 443, Remarks, Example 2, stata.toc
Replace the leading "d" on the sixth line of stata.toc with "t".

d ir irrefutable inference ... t ir irrefutable inference ...
(1) [R] ologit, p. 473, Syntax for predict
The option outcome(outcome) should be added to the syntax diagram for predict.

... nooffset ] ... outcome(outcome) nooffset ]

Also, a note should be added below the syntax diagram for predict that says:

Note that with the p option, you specify either one or k new variables depending upon whether the outcome() option is also specified (where k is the number of categories of depvar). With xb and stdp, one new variable is specified.
(1) [R] ologit, p. 474, Options, score(newvarlist) option
... would be named sc0, sc1, ... sck. ... would be named sc0, sc1, ... sc(k-1).
(1) [R] ologit, p. 474, Options for predict
Add the following option:

outcome(outcome) specifies for which outcome the predicted probabilities are to be calculated. outcome() should either contain a single value of the dependent variable, or one of #1, #2, ..., with #1 meaning the first category of the dependent variable, #2 the second category, etc.

Reference Manual, Volume 3

(all) [R] pergram, p. 20, Methods and Formulas, sixth paragraph, formula for f-hat
An i needs to be inserted after the pi symbol in the formula defining f-hat.
(all) [R] prais, p. 46, first paragraph, first sentence
... that maximizes the sum of squared residuals ... that minimizes the sum of squared residuals
(all) [R] regression diagnostics, p. 206, Methods and Formulas, second paragraph, second sentence
Mechanically, ... sum of squared residuals divided by 2. Mechanically, ... model sum of squares divided by 2.
(all) [R] reshape, p. 211, Example, output at bottom of page
The output for the command reshape wide inc ue, i(id) j(year) should show the reshaping going from long to wide.

... wide -> long ... long -> wide
(all) [R] signrank, p. 320, Methods and Formulas, signrank

Replace the following

With this distribution, the mean and variance of T are given by

E(T) = 0 and  Var_{adj}(T) = (1/4) \sum_{j=1}^n r_j^2

This variance is exact and is reported by signrank as the “adjusted variance”.

Note that the test statistic for the Wilcoxon signed-rank test is often expressed as the sum of the positive signed-ranks, but this just differs from T by a constant (that is fixed under the randomization distribution). Hence, T and the sum of the positive signed-ranks and the sum of the negative signed-ranks all have the same variance.

with

With this distribution, the mean and variance of T are given by

E(T) = 0 and  Var_{adj}(T) = \sum_{j=1}^n r_j^2

Note that the test statistic for the Wilcoxon signed-rank test is often expressed (equivalently) as the sum of the positive signed-ranks, T+, where

 E(T+)=n(n+1)/4 and  Var_{adj}(T+)=(1/4) \sum_{j=1}^n r_j^2

Within the following paragraph and in the three formulas after that (the fourth, third, and second formulas from the bottom of the page) replace all occurrences of (T) with (T+).

Also change the final formula on the page as follows.

 z = T/sqrt{Var_adj(T)}  z = (T+ - E(T+))/sqrt{Var_adj(T+)}
(1) [R] st stir, p. 415, Options, noshow option
prevents stsum ... prevents stir ...
(all) [R] st stir, p. 416, Saved Results, stir saved results list
r(p) two-sided p-value
r(p) one-sided p-value
(all) [R] st streg, p. 437, Distribution, table 1
For both the Weibull PH and Weibull AFT entries, change the survival function by removing the parentheses in the exponent expression.

 Bad Weibull Survival 
     Function  Good Weibull Survival 
     Function

For the Weibull AFT entry, change the Parameterization formula by removing the slash between the Beta and p in the exponent.

Bad Weibull Parameterization 
     Formula Good Weibull Parameterization 
     Formula
(1) [R] st streg, p. 440, Lognormal and log-logistic models, middle of page, formulas for h(t) and f(t)
The numerator for both the log–logistic hazard function h(t) and density function f(t) are the same and need to be changed as follows.

Bad h(t) & f(t) Numerator Good h(t) & f(t) Numerator
(all) [R] st streg, p. 440, Generalized gamma model, bottom of page, formula for f(t)
The three-parameter generalized gamma density function f(t) (both the not equal-to-zero part and the equal to zero part of the equation) need to be premultiplied by one over the product of sigma and t.
(1) [R] st sts list, p. 480, Options, noshow option
prevents sts graph ... prevents sts list ...
(all) [R] st stsplit, p. 540, Remarks, Example 3: Explanatory variables which change with time
Insert the following Stata command between the
. stsplit posttran, at(0,320)

and

. replace posttran=1 if posttran==320

commands.

. replace posttran=0 if tran==0

Also change the last command on the page as follows

. list id enter exit _t0 _t posttran ... . list id _t0 _t t1 posttran ...
(1) [R] st stvary, p. 554, Options, noshow option
prevents stsum ... prevents stvary ...

Reference Manual, Volume 4

(all) [R] svytab, p. 87, Saved Results, Matrices
e(prop)
e(obs)
e(deff)
e(deft)
e(Prop)
e(Obs)
e(Deff)
e(Deft)
(all) [R] svytest, p. 99, Methods and Formulas, second paragraph, second to last sentence
... computed using W/d ~ F(k,d). ... computed using W/k ~ F(k,d).
(all) [R] sw, p. 100, Syntax, the list of estimation commands supported
Add clogit and nbreg to the list of estimation commands supported by the sw command.
(1) [R] table, p. 144, Syntax
In the syntax diagram contents(clist) is optional and should appear within the brackets designating optional.

... , contents(clist) [ ... ... [, contents(clist) ...
(1) [R] table, p. 150, Options, contents(clist) option
contents(clist) is not optional; it specifies the contents of the table's cells. contents(clist) specifies the contents of the table's cells; if not specified, contents(freq) is used by default.
(1) [R] tsreport, p. 211, Syntax
The option report0 must be spelled out completely instead of allowing an abbreviation of r.
(1) [R] tsset, p. 215, Syntax
The option noxt should be deleted from the syntax diagram for tsset.

... [, noxt format(%fmt) ... ... [, format(%fmt) ...
(1) [R] tsset, p. 215, Options, noxt option
Option noxt and its description should be deleted.
(all) [R] xpose, p. 315, Description, second sentence
All new variables ... are made float. All new variables ... are made the default storage type (float unless reset by the set type command—see [R] generate).
(all) [R] xt, p. 317, Syntax
iis [varname_i] iis [varname_i] [, clear]
tis [varname_t] tis [varname_t] [, clear]
(all) [R] xt, p. 317, Options
Add the following paragraph describing the clear option.

clear removes the definition of i or t. For instance, typing iis, clear makes Stata forget the identity of the i() variable.
(all) [R] xtgee, p. 338, Syntax
Add the option trace to the xtgee syntax diagram.

... ] ... trace ]
(all) [R] xtgee, p. 340, Options
Add the following paragraph describing the trace option.

trace specifies that the current estimates should be printed at each iteration.
(1) [R] xtnbreg, pp. 383–390
Many changes were made to the xtnbreg manual entry in printing 2 of the Stata Reference Su–Z manual. Click here to obtain the second printing of this manual entry.
(all) [R] xtnbreg, p. 383, Syntax for predict
In addition to the changes in printing 2 (see the entry above), the syntax for predict for random-effects and conditional fixed-effects models should include the options nu0 and iru0.

... [, { xb | stdp } ... ... [, { xb | stdp | nu0 | iru0 } ...
(all) [R] xtnbreg, p. 384, Options for predict
In addition to the changes made in printing 2, add the following two paragraphs describing the nu0 and iru0 options.

nu0 calculates the predicted number of events assuming u_i = 0.
iru0 calculates the predicted incidence rate assuming u_i = 0.
(all) [R] xtpois, p. 391, Syntax for predict
The syntax for predict for random-effects and fixed-effects models should include the options nu0 and iru0.

... [, { xb | stdp } ... ... [, { xb | stdp | nu0 | iru0 } ...
(all) [R] xtpois, p. 393, Options for predict
Add the following two paragraphs describing the nu0 and iru0 options.

nu0 calculates the predicted number of events assuming u_i = 0.
iru0 calculates the predicted incidence rate assuming u_i = 0.
(1) [R] xtprobit, p. 406, Options for predict, xb option
... This is the default for the random-effects and fixed-effects models. ... This is the default for the random-effects model.
(all) [R] xtreg, p. 437, Methods and Formulas, xtreg, re, second paragraph
On the second line of the second paragraph of this section, make the following change.

 bad formula  good formula

On the third line of the second paragraph of this section, make the following change.

 bad formula  good formula
(all) [R] zip, p. 501, Syntax, Zero-inflated Poisson model
The option poisson should be deleted and the option vuong inserted in the syntax diagram for zip.

... offset(varname) poisson probit ... ... offset(varname) vuong probit ...
(all) [R] zip, p. 501, Syntax, Zero-inflated negative binomial model
The option nbreg should be deleted and the option vuong inserted in the syntax diagram for zinb.

... offset(varname) nbreg zip ... ... offset(varname) vuong zip ...
(all) [R] zip, p. 502, Options, poisson and nbreg options
The options poisson and nbreg and their descriptions should be deleted.
(all) [R] zip, p. 502, Options
Add the following paragraph describing the vuong option.

vuong specifies that the Vuong (1989) test of the zero-inflated Poisson model vs. the Poisson model (or zero-inflated negative binomial model vs. negative binomial model) be reported. This test statistic has a standard normal distribution with large positive values favoring the zip (zinb) model and large negative values favoring the Poisson (negative binomial) model.
(1) [R] zip, p. 502, Options, zip option
The last sentence in the zip option concerning the Wald test should be deleted.

zip ... The default is to include a Wald test. zip ...
(all) [R] zip, pp. 503–504, Example
The entire example starting in the middle of p. 503 and continuing through the middle of p. 504 has been replaced. Click here to view the correction.
(all) [R] zip, p. 507, References
Add the following entries to the References section.

Greene, W. H. 1994. Accounting for excess zeros and sample selection in Poisson and negative binomial regression models. Working paper no. EC-94-10, Department of Economics, Stern School of Business, New York University.

Vuong, Q. 1989. Likelihood ratio tests for model section and non-nested hypotheses. Econometrica 57: 307–334.

Also update the Greene (1997) reference to the fourth edition.

Greene, W. H. 1999. Econometric Analysis, 4th ed. Upper Saddle River, NJ: Prentice–Hall.

User’s Guide

(all) [U] 1.3.3 Changes to existing commands, p. 14, second paragraph
brier ... test of the ROC area being greater than .5; ... brier ... test that a brier score is extreme; ...
(all) [U] 1.3.6 New programming features, p. 20, second paragraph under "Other matrix changes:"
You can ... instead refer to A[i,...]. You can ... instead refer to A[i,1...].
(all) [U] 16.3.4 Time-series functions, p. 126, monthly() entry
monthly(s1,s2[,y)]) monthly(s1,s2[,y])
(all) [U] 17.9 Subscripting, p. 160, last line of item 3
generate hat = price*b[1,"mpg:price"] generate hat = price*b[1,colnumb(b,"mpg:price")]
(all) [U] 21.3.5 Double quotes, p. 194, seventh line
Why is `"example'" better than ... Why is `"example"' better than ...
(all) [U] 27.3.1 Inputting time variables, p. 294, sentence before the table
We ... for %d variables; ... We ... for %t variables; ...
(all) [U] 27.3.6 Creating time variables, p. 297, command after first paragraph
. generate time = d(19100q1)+_n-1 . generate time = q(19100q1)+_n-1