CONTENTS

INTRODUCTION

STATA SYNTAX

DATASET FILES
OPENING A DATASET
FROM EXCEL TO STATA
WORKING WITH LARGE DATASETS
SAVING DATA
COMBINING DATASETS
EXPLORING THE DATA

SUMMARIZING DATA

GENERATING VARIABLES
THE GENERATE COMMAND
CREATING DUMMY VARIABLES

REGRESSIONS
LINEAR REGRESSION
PROBIT MODELS

WRITING A PROGRAM: DO-FILES

RECORDING OUTPUT: LOG FILES

MORE HELP

GOOD COMMANDS TO KNOW

INTRODUCTION

Stata is a powerful statistical software package, used by students and researchers in many fields.
Stata can manipulate data, calculate statistics, and run regressions. You will to need to use Stata
to complete problem sets and write research papers for your economics classes. This manual is
an introduction to Stata’s basic features, covering the applications and functions that you are
likely to use most often. It will explain the three types of files Stata uses: datasets (.dta), do-files
(.do), and log files (.smcl).

The examples provided in this guide use the dataset wages.dta, which is available along with
this manual and a sample do-file at the Economics Department website.

Stata is available on a number of computers on campus. To see a full list of computers that have
Stata, go to the LIS website, select "Quick Links" and then click "Software on Library & Lab
Computers." As of the writing of this manual, Stata was available in at least some labs in
Sunderland, Munroe, the Library, and Bi-Hall.

When you first open Stata, you will see a screen that looks like this:

The large Results window shows output. Commands are typed into the Command window
below it. The Review window displays recently executed commands – to repeat a command
without typing it again, you can click on it in this window. The Variables window displays the
names and labels of the variables in your dataset.

An Introduction to Stata 1
If you place the cursor over buttons in the toolbar, you will see that they are, from left to right,
Open..., Save, Print, Begin Log, View..., Graph, Do-File Editor, Data Editor, Data Browser,
Go, and Break.

STATA SYNTAX

You will need to input commands using Stata syntax. Stata is caps-sensitive; commands do not
use capital letters and variable names should only be inputted with capital letters if that is how
they appear in the variable window.

As we go through some Stata commands in this manual, you will notice that most follow a basic
syntax:

command varlist if exp, options

The command tells Stata what action to take.

The varlist tells Stata what variables to take this action on. This is optional in many
cases.

The if exp expression limits the command to a selected sample of the data set. This is
always optional.

The , options allows you to add options that are specific to that command.

An example that uses all elements of this syntax appears in the "Summarizing Data"
section.

Commands can generally be abbreviated to save typing. You don’t need to type anything more
than su in place of summarize or ta for tabulate, but you need to type out the whole word for
tabstat.

There is a list of common commands to familiarize yourself with at the end of this guide.

An Introduction to Stata 2
DATASET FILES

OPENING A DATASET
Datasets in Stata format have the extension .dta. There are several ways to open a .dta file in
Stata:

• Click on the .dta file. This will open Stata and load the data in a new window.

• Open Stata. From the File menu, choose “Open” and select the dataset.

• Open Stata. Type use filename.

Note: In order for this to work, you must be running Stata in the same directory
(folder) as the file you need to open. To see what directory you are in, type pwd.
To change the directory, type cd, followed by the name of the new directory. If
your files are on Tigercat, type cd U:\

If you do not wish to change directories, you can type out the entire file path, that
is, type use "U:\wages.dta" if the file is on Tigercat.

The third method seems the most complicated at first, but it is the best method to use when
working with do-files since the command can be included in the do-file, allowing you to avoid
having to open the file manually each time before running the do-file.

FROM EXCEL TO STATA

Often, data for a project will be in a different format, like an Excel spreadsheet. Usually, data in
Excel requires some editing before it can be transferred to Stata. First, make sure that the first
row of the Excel sheet is the variable names and that the observations begin in the next row. If
there are no variable names, then the observations can start in the first row, but there cannot be
any extra rows. Any missing values should be an empty cell, not a space or a dot. Make sure
that the variable names do not contain any spaces. This is also a good time to make sure that the
variable names make sense. Concise names that make sense to you are best. You can change the
names in Stata if you wish to, but many people find it easier to do in Excel.

The easiest way to transfer a small Excel data set into Stata is to use copy and paste. Make sure
that your excel spreadsheet is ready to copy. Open Stata and open the Data editor, which will
look similar to a blank spreadsheet. Then, select all of the cells in the Excel file, copy them, and
paste them into Stata’s data editor.

An alternative is to save the Excel data as a tab-delimited text file. (File>Save As… In the
“Save as type” dropdown menu, choose “Text (Tab delimited) (*.txt).) Then, in Stata, type
insheet using filename.txt. (Again, you must be running Stata in the directory where you saved

An Introduction to Stata 3
the .txt file. By now you may have noticed that it is best to change the directory to Tigercat
when you open Stata.) This imports the dataset into Stata. You can type browse or click the
Data Browser button in the toolbar (it has a magnifying glass) to look at the data. Look for
obvious mistakes. Any observations in red are variables that Stata has interpreted as string, or
nonnumeric variables. You won’t be able to run regressions with these. If a variable shouldn’t
be a string variable, try to figure out why Stata interpreted it that way (e.g., one of the
observations for that variable includes something like “na” for not available instead of a
number), fix it in Excel, and input the data into Stata again.

WORKING WITH LARGE DATASETS

If a dataset is very large, you may receive an error message when you try to open it in Stata (“no
room to add more variables”). In this case, you can increase the memory available to Stata by
typing:
clear
set memory Xm
where X is the amount of memory you need. The default setting is 10m (MB). If your dataset is
larger than 10MB, Stata will not be able to open it. To find out how much memory your dataset
will need, right-click on the file and select Properties... Set the memory in Stata to be slightly
larger than the size of the file. After you set the memory, try to open the dataset again.

Once you are able to load the data, the command compress may reduce the size of the dataset
without changing the data. The variable may be formatted to use, for example, 8 digits, but if no
observation uses more than two, Stata will eliminate the extra places to make space. This only
works if there are extra places to eliminate.

SAVING DATA
Stata does not automatically save changes you make to the data. To save your data, use the save
command. Typing save filename, where you choose any filename, will save your data as
filename.dta in the current directory. If you have already used the cd command to change the
directory to Tigercat, then you can type save filename to save to Tigercat. Otherwise, you will
have to type save "U:\filename" to save the file to Tigercat.

If you have already used the file name, then typing save, replace will replace the old file with
the new version. Stata will not overwrite the old data unless you type out replace. Otherwise, if
you try to use an existing file name, you will see an error message – this makes it difficult to
accidentally overwrite data. If you would like to keep the old version of the data, choose a new
file name.

It’s a good idea to keep a clean copy of the original dataset so that you can start over if, for
example, you accidentally drop a crucial variable.

An Introduction to Stata 4
COMBINING DATASETS

Instead of just one dataset, you might have data from several sources that you need to combine.
If you want to stack data sets—that is, add more observations to the same variables—you want to
combine data vertically. This situation might arise if, for instance, you had separate data sets for
2005 and 2006 and wanted to stack them into a single data set. If instead you want to add new
information – new variables – about the same observations, you want to combine the data
“horizontally.”

Visual representation of combining data sets A and B

A B

Append Merge

Combining data “vertically” is done using the append command. Both datasets need to be in
Stata format (they need to be .dta files). Open the first dataset in Stata. Then, using the
command append using seconddata.dta, where seconddata is the name of the second dataset,
will combine the datasets.

To combine data “horizontally,” use the command merge. Using merge is trickier than using
append, because you have to ensure that the observations in each dataset correspond to each
other. Usually you will want to merge data based on some kind of identification number
(individual, household, etc). Both datasets need to be saved as .dta files. To merge the data,

An Introduction to Stata 5
open the first dataset. Type sort id (where id is the name of your identification variable), then
save, replace. Type clear and then open the second dataset. Type sort id again, and save the
second dataset. While this second dataset is open, type merge id using firstdata.dta (where
firstdata is the name of the first dataset) to combine the datasets. This performs what Stata calls
a match merge. Note that this command will not work if the identification numbers are not
unique (there must only be one observation per identification number).

EXPLORING THE DATA

Once you have your dataset open, you may wish to use Stata to familiarize yourself with it. You
could begin by opening the data editor, which will show you a spreadsheet-like representation of
your data. However, if your data sets has many variables and/or observations, this will be an
unwieldy way to look at it. You may instead wish to use the describe command, which will
return basic information about your data set and a list of variables. The command lookfor x is
also helpful; it returns any variables that contains x in the name or description. For instance, if
you were looking for a variable related to age, you might type lookfor age to see what variables
have age in the name or description.

An Introduction to Stata 6
SUMMARIZING DATA

The first type of analysis that many people run when their data are ready is basic summary
statistics. Remember that the Stata syntax of most commands is command varlist if exp,
options. The following examples will take you through obtaining summary statistics using this
syntax and examples of output from the sample dataset wages.dta.

summarize tells Stata to list the number of observations, mean, standard deviation, minimum,
and maximum for each variable. This is the most basic syntax: it is only a command with no
options.

Example
. summarize

Variable Obs Mean Std. Dev. Min Max

id 0
age 200 44.01 17.89023 16 85
sex 200 1.585 .4939585 1 2
education 200 39.875 3.030726 32 46
race 200 1.49 1.536785 1 18

ethnic 22 1.954545 1.557693 1 5

wkly_hours 115 38.67826 8.047608 10 66
wkly_earn 123 678.1258 524.3861 0 2884.61
region 200 2.62 1.077686 1 4

The varlist, or variable list, tells Stata which variables to apply the command to. Type the
names of the variables you want to use (or click on the names in the variables window to the left
of the command window) without commas.

Example

summarize wkly_earn tells Stata to summarize weekly earnings only.

. summarize wkly_earn
Variable Obs Mean Std. Dev. Min Max

wkly_earn 123 678.1258 524.3861 0 2884.61

Using if exp, short for “if expression”, limits the command to certain values of a particular
variable.

In these expressions:

An Introduction to Stata 7
== means “is equal to” Note that you must use two equal signs because this is a logical
statement rather than a mathematical formula
!= means “does not equal”
~= also means "does not equal"
> means “greater than”
< means “less than”
>= means “greater than or equal to”
<= means “less than or equal to”
& means "and"
| means "or"

Note: Stata reads a missing value as infinity, so be careful when using these expressions.
Stata might perform a command on missing observations that you intended to leave out.

Examples

summarize wkly_earn if sex==1 summarizes the weekly earnings of men as indicated

by sex taking on a value of 1.

. summarize wkly_earn if sex==1

Variable Obs Mean Std. Dev. Min Max

wkly_earn 54 746.4437 575.5715 0 2884.61

summarize wkly_earn if wkly_hours>=40 & sex==2 reports the number of

observations that are women and work full time (40 or more hours per week).

. summarize wkly_earn if wkly_hours<40 & sex==2

Variable Obs Mean Std. Dev. Min Max

wkly_earn 12 373.5792 292.7015 100 1080

summarize wkly_earn if sex==1|sex==2 reports the number of observations that are

male or female. Of course, this is everyone in the sample.

. summarize wkly_earn if sex==1|sex==2

Variable Obs Mean Std. Dev. Min Max

wkly_earn 123 678.1258 524.3861 0 2884.61

An Introduction to Stata 8
The modifier bysort can precede a command.

Example

bysort sex: summarize wkly_earn summarizes weekly earnings separately for each
value of sex, that is, for men and for women.
. bysort sex: summarize wkly_earn

-> sex = 1

Variable Obs Mean Std. Dev. Min Max

wkly_earn 54 746.4437 575.5715 0 2884.61

-> sex = 2

Variable Obs Mean Std. Dev. Min Max

wkly_earn 69 624.6596 478.0536 0 2403

You can also include options with certain commands. Options come last in the command syntax
and are preceded by a comma.

Example

Adding the option detail to the command summarize provides additional summary
statistics.
. summarize wkly_earn, detail
Earnings per week
Percentiles Smallest
1% 0 0
5% 100 0
10% 220 51 Obs 123
25% 346 82.4 Sum of Wgt. 123
50% 560 Mean 678.1258
Largest Std. Dev. 524.3861
75% 880 1923.07
90% 1403.84 2403 Variance 274980.8
95% 1680 2884.61 Skewness 1.89694
99% 2884.61 2884.61 Kurtosis 7.517251

An Introduction to Stata 9
The previous examples have all used the summarize command. Here are some other helpful
commands for generating summary statistics.

tabulate makes a frequency table.

. tab sex
Sex: 1 =
male, 2 =
female Freq. Percent Cum.
1 83 41.50 41.50
2 117 58.50 100.00
Total 200 100.00

codebook lists information about the values a variable takes. It shows the range, common
percentiles, and how many unique values a variable has. It also shows how many observations
are missing.
. codebook wage

wage (unlabeled)

type: numeric (float)

range: [0,72.11525] units: 1.000e-06
unique values: 82 missing .: 89/200
mean: 17.8266
std. dev: 12.3417
percentiles: 10% 25% 50% 75% 90%
8 10 14.82 20.3846 30

tabstat allows you to customize a table of summary statistics.

Example

tabstat varlist, by(variable) stat(range sd p25 p50 p75) creates a table, sorted by
variable, that includes the range, standard deviation, and 25th, 50th, and 75th percentiles of
the specified variables. There are dozens of statistics options, see the online Stata help
for tabstat for a complete list.

. tabstat wage, by(sex) stat(sd p25 p50 p75 range)

Summary for variables: wage

by categories of: sex (Sex: 1 = male, 2 = female)

sex sd p25 p50 p75 range

1 13.80836 10.15 16.34324 24 72.11525
2 10.9587 9.6 14 19 54.405

Total 12.34169 10 14.82 20.3846 72.11525

An Introduction to Stata 10
GENERATING VARIABLES

THE GENERATE COMMAND

In Stata, you can generate a new variable using the command generate. The general syntax is
generate newvarname = expression. You can include constants, variables, or both in the
expression.

When using functions to generate a new variable,

* means multiply
/ means divide
+ means add
- means subtract
^ makes an exponent
ln(var) or log(var) takes the natural log of var

Example

Using the sample data, suppose you want to create an hourly wage variable by dividing
weekly earnings by hours worked. Type generate wage = wkly_earn/wkly_hours.
. generate wage = wkly_earn/wkly_hours
(89 missing values generated)

Notice that Stata returns a message that 89 missing values have been generated. If Stata
is unable to generate an observation, in this case because either wkly_earn or
wkly_hours is missing, or wkly_hours is equal to zero, it records a missing value and
notifies you.

If you wanted to include log wage in a regression, you could then generate a log wage
variable by using the command generate lwage = ln(wage). The missing values
generated here result from a missing value for wage or a value of 0 for wage (since the
natural log of 0 is undefined).
. generate lwage = ln(wage)
(90 missing values generated)

GENERATING DUMMY VARIABLES

You may often need to generate dummy variables to use in a regression. Dummy variables (also
commonly called "indicator variables" or "binary variables") take on two values: 0 and 1.You
can create a dummy variable using either a single command or two commands. Both options are
shown in the examples below.

An Introduction to Stata 11
Example: Generating a dummy variable based on a continuous variable
Suppose you need to generate a dummy to indicate that an individual is of retirement age.
age is a continuous variable, and you want your dummy variable to indicate when
someone is over 65.

Option 1: Single command

generate retire = age>65

This creates a variable named retire that is equal to 1 if the expression is true (if the
value of age is greater than 65) and equal to 0 if the expression is not true. By tabbing
the new dummy variable, we see that 26 of the 200 observations (13%) are of retirement
age.
. generate retire = age>65

. tab retire

retire Freq. Percent Cum.

0 174 87.00 87.00

1 26 13.00 100.00

Total 200 100.00

Option 2: Two commands

generate retire=1 if age>65
replace retire=0 if age<=65
This creates the same variable as in the first option, but you are being more explicit about
what the conditions are. The first command generates retire, which is equal to 1 for
people over 65 and is missing for everyone else. The second command replaces retire
with 0 for people who are 65 or under, changing those missing values to zeros.

. generate retire=1 if age>65

(174 missing values generated)

. tab retire
retire Freq. Percent Cum.

1 26 100.00 100.00
Total 26 100.00

. replace retire=0 if age<=65

(174 real changes made)
. tab retire

retire Freq. Percent Cum.

0 174 87.00 87.00
1 26 13.00 100.00

Total 200 100.00

An Introduction to Stata 12
The above example works because the age variable in the sample dataset has no missing values.
If any values were missing, Stata would code them as 1 in the new variable, because a missing
value is read as infinity, which of course is greater than 65. I would be able to fix this by using
the command replace retire=. if age==. after generating the variable.

The previous example involved creating a dummy variable based on a continuous variable, age.
Data often includes categorical variables, which are variables that take on numeric values that
are not inherently meaningful but instead correspond to different categories (region, race, and
sex are examples variables that are often coded this way). A codebook usually accompanies the
data to explain how the variables are coded, and in the case of a categorical variable, what each
variable means. To include such a variable in a regression, you will need to create dummy
variables for each category. The following examples cover several common situations that arise
when creating dummy variables based on categorical variables.

Example: Generating a dummy variable based on a categorical variable that takes on two
values

In the sample dataset, sex is set equal to 1 if an individual is male and 2 if she is female.
So that the mean and coefficient for this variable can be easily interpreted, it should be
converted to a dummy variable. The command generate male= sex==1 creates a dummy
variable equal to 1 for males and 0 for females.
. tab sex

Sex: 1 =
male, 2 =
female Freq. Percent Cum.

1 83 41.50 41.50
2 117 58.50 100.00

Total 200 100.00

. generate male = sex==1

. tab male

male Freq. Percent Cum.

0 117 58.50 58.50

1 83 41.50 100.00

Total 200 100.00

Note: The same warning about missing values in the retire example above applies to the
creation of dummy variables based on a categorical variable. If there are any missing
observations in the original variable they will be coded as either 0s or 1s, depending on how you
specify the command. If your variable does have missing observations, you will need to type
replace newvar = . if oldvar==. after generating the variable to ensure that any observations

An Introduction to Stata 13
that were missing remain missing in the new dummy variable. In this example, sex does not
have any missing observations, so there is no problem to correct.

Example: Generating dummy variables based on a categorical variable that takes on

more than two values

In the sample dataset, region is a categorical variable coded as 1 if a household is in the

northeast, 2 if it is in the midwest, 3 if it is in the south, and 4 if it is in the west. If you
wanted to control for region in a regression, you would need to create three dummy
variables (because the fourth is omitted). This example does so using a different method
for each variable.

. tab region

1 =
northeast,
2 =
midwest, 3
= south, 4
= west Freq. Percent Cum.
1 42 21.00 21.00
2 42 21.00 42.00
3 66 33.00 75.00
4 50 25.00 100.00

Total 200 100.00

To create a dummy variable called northeast, equal to 1 if the individual lives in the
northeast and 0 if he does not, I can type generate northeast=1 if region==1. This
creates 1s for all the observations corresponding to individuals in the northeast, but
observations corresponding to individuals outside of the northeast are created as missing
values. I can replace the missing values with 0s by typing replace northeast=0 if
region!=1.

. generate northeast=1 if region==1

(158 missing values generated)

. replace northeast=0 if region!=1

(158 real changes made)

. tab northeast

northeast Freq. Percent Cum.

0 158 79.00 79.00
1 42 21.00 100.00

Total 200 100.00

An Introduction to Stata 14
I could also simply type generate midwest=0, which creates a variable called midwest
and sets every observation equal to 0, and then replace midwest=1 if region==2 to
change the 0s to 1s for those households in the midwest.
. generate midwest=0

. replace midwest=1 if region==2

(42 real changes made)

. tab midwest
midwest Freq. Percent Cum.

0 158 79.00 79.00

1 42 21.00 100.00

Total 200 100.00

An even easier way is to type generate south= region==3. This makes the 0s and 1s in
one step.

. generate south= region==3

. tab south
south Freq. Percent Cum.

0 134 67.00 67.00

1 66 33.00 100.00

Total 200 100.00

Example: Generating dummy variables that incorporate multiple values of a categorical

variable

A categorical variable can take on many values, but you might not need or want to create
a separate dummy variable for each possible value of the categorical variable. The
education variable in the sample dataset is an example. The values in education
correspond to the highest grade an individual has completed. Instead, of making a
separate dummy variable for each different grade, suppose you want to create only three
dummy variables: one for individuals who did not finish high school, one for individuals
who have a high school diploma but no college degree, and one for individuals who have
some kind of college degree.

The education variable looks like this:

An Introduction to Stata 15
 . tab education

Highest
grade
attended Freq. Percent Cum.
32 3 1.50 1.50
33 6 3.00 4.50
34 4 2.00 6.50
35 5 2.50 9.00
36 11 5.50 14.50
37 9 4.50 19.00
38 3 1.50 20.50
39 52 26.00 46.50
40 35 17.50 64.00
41 8 4.00 68.00
42 8 4.00 72.00
43 38 19.00 91.00
44 11 5.50 96.50
45 5 2.50 99.00
46 2 1.00 100.00

Total 200 100.00

The codebook that came with the data explains that the number 32 indicates less than a
first-grade education, 33 indicates that an individual has completed at least the first grade
but not beyond the fourth grade, 34 indicates that an individual has completed the fifth or
sixth grade, and so on. Suppose that you wish to use this variable to create dummy
variables for three mutually exclusive and collectively exhaustive educational categories:
less than high school, high school diploma, and beyond high school. You note that you
want to code these dummy variables using education as follows:
. tab education

Highest
grade
attended Freq. Percent Cum.
32 3 1.50 1.50
33 6 3.00 4.50
34 4 2.00 6.50
no diploma
35 5 2.50 9.00
36 11 5.50 14.50
37 9 4.50 19.00
38 3 1.50 20.50
high school 39 52 26.00 46.50
40 35 17.50 64.00
41 8 4.00 68.00
42 8 4.00 72.00
college 43 38 19.00 91.00
44 11 5.50 96.50
45 5 2.50 99.00
46 2 1.00 100.00

Total 200 100.00

To generate a dummy variable for no high school diploma, type generate nodiploma=1
if education<=38 then replace nodiploma=0 if education>38.

An Introduction to Stata 16
 . generate nodiploma=1 if education<=38
(159 missing values generated)

. replace nodiploma=0 if education>38

(159 real changes made)

. tab nodiploma

nodiploma Freq. Percent Cum.

0 159 79.50 79.50
1 41 20.50 100.00

Total 200 100.00

To generate a dummy variable for those who have a high school diploma but no college
degree, we need to include categories 39 (high school diploma) and 40 (some college but
no degree). Type generate hsdiploma=0, then replace hsdiploma=1 if education>=39
& education<=40.
. generate hsdiploma=0

. replace hsdiploma=1 if education>=39 & education<=40

(87 real changes made)

. tab hsdiploma
hsdiploma Freq. Percent Cum.

0 113 56.50 56.50

1 87 43.50 100.00

Total 200 100.00

The remaining individuals have some kind of college degree. Use the command
generate college= education>40 to generate a dummy variable for this category.

. generate college= education>40

. tab college

college Freq. Percent Cum.

0 128 64.00 64.00

1 72 36.00 100.00

Total 200 100.00

An Introduction to Stata 17
REGRESSIONS
Once your data are in Stata format and you have generated any necessary variables, you can
begin to run your regressions.

LINEAR REGRESSION
The command for a ordinary least squares (OLS) regression is simply regress followed by the
variables you are using, with the dependent variable (what you are predicting, the y variable)
listed first, followed by any independent variables: regress yvar xvar1 xvar2 …

Using the sample dataset, I could run a regression to look for a relationship between age and
lwage, the log wage variable that I generated earlier.

. regress lwage age

Source SS df MS Number of obs = 110

F( 1, 108) = 2.60
Model .791520315 1 .791520315 Prob > F = 0.1096
Residual 32.8370879 108 .30404711 R-squared = 0.0235
Adj R-squared = 0.0145
Total 33.6286082 109 .308519342 Root MSE = .5514

lwage Coef. Std. Err. t P>|t| [95% Conf. Interval]

age .0060651 .003759 1.61 0.110 -.001386 .0135161

_cons 2.473424 .1625819 15.21 0.000 2.151158 2.795689

I can add more x-variables to the regression:

. regress lwage male age hsdiploma college

Source SS df MS Number of obs = 110
F( 4, 105) = 13.80
Model 11.5855578 4 2.89638946 Prob > F = 0.0000
Residual 22.0430504 105 .209933813 R-squared = 0.3445
Adj R-squared = 0.3195
Total 33.6286082 109 .308519342 Root MSE = .45819

lwage Coef. Std. Err. t P>|t| [95% Conf. Interval]

male .0973496 .0905015 1.08 0.285 -.0820982 .2767974
age .0063431 .0031265 2.03 0.045 .0001438 .0125423
hsdiploma .111912 .1494281 0.75 0.456 -.1843764 .4082004
college .7094032 .1498775 4.73 0.000 .4122237 1.006583
_cons 2.079982 .1976793 10.52 0.000 1.688021 2.471944

Using the option robust will display robust standard errors in the regression output.

An Introduction to Stata 18
 . regress lwage male age hsdiploma college, robust
Linear regression Number of obs = 110
F( 4, 105) = 15.26
Prob > F = 0.0000
R-squared = 0.3445
Root MSE = .45819

Robust
lwage Coef. Std. Err. t P>|t| [95% Conf. Interval]

male .0973496 .0932892 1.04 0.299 -.0876256 .2823248

age .0063431 .0028214 2.25 0.027 .0007486 .0119375
hsdiploma .111912 .1311612 0.85 0.395 -.1481565 .3719805
college .7094032 .1326931 5.35 0.000 .4462972 .9725091
_cons 2.079982 .1789866 11.62 0.000 1.725085 2.434879

PROBIT MODELS
To run another type of regression, like a probit, the syntax is the same: probit yvar xvar1
xvar2… Although the coefficients in a probit model do not indicate marginal effects, typing mfx
directly after running the regression will display a table of the marginal effects.
. probit employed age male

Iteration 0: log likelihood = -136.37092

Iteration 1: log likelihood = -131.24364
Iteration 2: log likelihood = -131.23857

Probit regression Number of obs = 200

LR chi2(2) = 10.26
Prob > chi2 = 0.0059
Log likelihood = -131.23857 Pseudo R2 = 0.0376

employed Coef. Std. Err. z P>|z| [95% Conf. Interval]

age -.0149433 .0051216 -2.92 0.004 -.0249814 -.0049052

male .1755545 .1849013 0.95 0.342 -.1868452 .5379543
_cons .7760095 .2601069 2.98 0.003 .2662093 1.28581

. mfx
Marginal effects after probit
y = Pr(employed) (predict)
= .57581905

variable dy/dx Std. Err. z P>|z| [ 95% C.I. ] X

age -.0058535 .00201 -2.91 0.004 -.009792 -.001915 44.01
male* .0684797 .07172 0.95 0.340 -.072098 .209058 .415
(*) dy/dx is for discrete change of dummy variable from 0 to 1

You could also have obtained the marginal effects in a single command by entering
dprobit employed age male.

An Introduction to Stata 19
DO-FILES

Do-files are programs that you write with all of the commands that you wish to run for your
analysis. They have the extension .do.

When you type commands into the command window one at a time, you are working
interactively. Creating a do-file is a very useful alternative to working interactively. Do-files
allow you (or your professor or a Stata tutor) to easily reproduce your work. Saving your work
in a do-file will let you pick up where you left off the last time you used Stata or perform the
same set of operations on multiple datasets. Most importantly, if you realize that you made a
mistake or forget a command at the beginning of a session, you can edit the do-file and run it
again instead of manually re-entering each individual command into the command window. If
you get stuck, a Stata tutor can look at your do-file to find out what went wrong.

You should not undertake a research project (or even a long homework) without writing a
do-file.

To start a do-file, click on the “New Do-file Editor” button in the Stata toolbar (it looks like a
notepad), or select This opens a blank window with a cursor. Type commands into the window,
on separate lines, in the order you want Stata to perform them. You can make notes in your do-
file by beginning a line with an asterisk (*).

Save your do-file by using the File menu or the Save button in the do-file editor.

There are two ways to carry out the commands in a do-file:

• Run performs all the commands in the do-file but does not show any output. If there are
commands in the do-file that change the data, the changes will be made.
• Clicking do performs all the commands in the do-file and shows output in the results
window for each command, just the same as if you had typed the commands into the
command window. Typing save into the do-file saves your data, not the do-file itself.

You can run or do a do-file by clicking the buttons on the right of the toolbar in the do-file
editor.

To open an existing do-file, right-click on the file and select Edit. This opens the do-file editor.
Alternatively, open the do-file editor and select Open… from the File menu.

It is also possible to create and edit do files in any other text editor. To invoke them one simply
types “do filename.do” on the command line.

An Introduction to Stata 20
RECORDING OUTPUT: LOG FILES

To avoid losing the results of analyses carried out each time you quit Stata, you will have to
create a log file. Stata’s log files have the extension .smcl. The command to begin a log file is
log using logname. This creates a new file, logname.smcl, in the current directory, in which
everything that was seen in the results window during that session will be recorded. You could
also select Log > Begin… from the File menu to open a log. When you are finished, you type
log close or just exit Stata to close the log; your output will be saved automatically. The next
time you open Stata, you can create a new log with a new name. If you would prefer to keep all
your output in one log file, type log using logname, append, which will append the new log to
the end of the existing file. To replace the old log with the new one, type log using logname,
replace. As with any command, these can be incorporated into your do file so that it is
automatically done each time you run the program.

To view a log file, select Log > View… from the File menu and choose the log file, or right-
click on the file and select Open. This opens Stata’s log viewer. You can print a log file from
the log viewer to attach it to a problem set. Results can also be copied from the log viewer or
from the results window in Stata and pasted into a problem set in Microsoft Word. Tables from
Stata do not usually align properly in Times or Times New Roman, but changing the font to
Courier or Courier New in size 9 or 10 will fix this. Since even this doesn’t look very
professional, for a formal research paper the best method is to create a neatly formatted results
table using Excel.

MORE HELP

Stata official help website is extensive. You can get there by typing search followed by your
problem into Stata, or through Google. If that doesn’t solve your problem, consulting the Stata
manuals can help. The manuals have more detail than the online help and include many useful
examples. You can check out the manuals at the library. A set of manuals also is available in
Sunderland ILC1; these manuals cannot be removed from the lab. Please be respectful to others
and leave them there.

The Economics Department also has Stata tutors who are available to help you. These tutors will
hold training sessions at the beginning of the Fall and Spring semesters and and can also be
consulted during their office hours. Check the department website for a list of tutors and
scheduled office hours.

An Introduction to Stata 21
GOOD COMMANDS TO KNOW

(This list is NOT a comprehensive list of commands that Stata uses, but these are the commands
you may use most frequently.)

MANAGING DATA
pwd shows which directory Stata is in

cd U:\ changes the directory to Tigercat

use "filename.dta" opens a dataset

set mem Xm sets the memory in Stata to X MB

compress may reduce the size of a dataset by storing the data more efficiently

insheet imports a tab-delimited (or comma-delimited) text file into Stata

browse opens the data browser

edit opens the data editor so you can manually edit the data

drop removes specified variables or observations from the dataset

keep command keeps the specified variables or observations and removes the rest

sort orders the data in a certain way. Typing sort age would order the observations from
smallest age to largest age.

save saves your data, a common option is replace

append combines two datasets “vertically,” adding observations to variables

merge combines datasets “horizontally”

rename oldname newname changes a variable name from oldname to newname

label adds a descriptive label to a variable. Type label var varname “label”

generate creates a new variable

The command egen also creates variables. egen is usually used for creating variables that have
something to do with summary statistics

An Introduction to Stata 22
describe lists variables and their labels.

replace replaces specified observations with a value you can specify.

SUMMARIZING DATA
summarize provides basic summary statistics

tabulate makes a frequency table.

codebook lists information about the values a variable takes

tabstat allows you to customize a table of summary statistics.

REGRESSIONS
regress runs an ordinary linear regression

probit runs a probit model

mfx displays marginal effects for the most recent regression

xtreg performs fixed or random-effects linear regression (for panel data).

An Introduction to Stata 23