Introduction to Stata
LPO 9951 | Fall 2015
PURPOSE
In this class we’ll walk through some of Stata’s basic functionality. We’ll also get used to the idea of interacting with Stata through the command line and *.do files.
Stata as a calculator
Stata can be used as a calculator via the display command. All of the normal rules of arithmetical precedence apply to the Stata syntax.
. display sqrt(42)
6.4807407
. di sqrt(42) + 4
10.480741
. di (sqrt(42) + 4) - 10
.4807407NB: The display command can be shortened to di. Many Stata commands and options are like this. The help files underline the minimum part of the command/option that must be specified in order for the package to understand what you want.
Using *.do files
Stata syntax is stored in what’s called *.do files. These are all of the typed commands that you use to manipulate and analyze the data. A properly formatted *.do file can be run from the command line using the do command:
. do "../do/lecture1_introduction_hello.do"
. display "Hello, World!"
Hello, World!
end of do-fileOne of the key skills you’ll learn this year is properly annotating a *.do file. Remember that these files are primarily meant to be read by humans, and only incidentally meant to be read by computers. Stata assumes that everything in a *.do file is a command unless it’s preceded by a comment sign. To set off a line of text as a comment, place * or // in front of it. You can also use the /* ... */ format for comments, which can be used on the same line as the syntax itself:
. * comments can be on their own rows...
. // ...like this
. /* ...and this, or */
. di 1 // to the
1
. di 2 /* side of commands */
2Directory structure
We’ll talk about directory structure more detail later, but for now, make sure that your course files have at least the following structure:
.
|-- /data
| |
| |-- <data files>
|
|-- /do
| |
| |-- <Stata do files>
|
|-- /plots
| |
| |-- <plot files>Place your Stata do files in the ./do subdirectory, all data files in the ./data folder, and all saved graphics in the ./plots. We’ll add more as the semester goes on but these will do for now. The primary directory (represented by the .) can be anywhere on your computer or a thumbdrive. What really matters are the relative paths between the subfolders. Just make sure that wherever you choose to hold your course files you have enough storage space. While do files are usually very small, some of our datasets will be fairly large.
Loading Stata data files
All Stata data files are saved in the *.dta format. Today we’ll be using the census.dta file which contains information on characteristics of the 50 states from the 1980 census.
To locate a data file, you first have to tell Stata where to look on your computer. With some very rare exceptions, you should always use the cd command to set the working directory:
. cd "~/Github/lpo9951/markdown"
/Users/benski/Github/lpo9951/markdownNB: The exception to this rule would be (1) when you double-click your *.do file and have Stata configured to open automatically; and (2) your *.do file is set to work in the directory in which it is currently located (i.e., all the relative links are correctly specified).
The above directory is where I keep the class files, hence the cd command doesn’t really do anything. Your files will be in a different location on your computer. Changing the working directory just once makes it much easier to exchange *.do files across computers. Don’t place a cd command in your *.do file. This will make collaboration much easier.
To open a Stata file, use the use command:
. use "../data/census.dta", clear
(1980 Census data by state)
. We’ll go over for other commands for importing more complex data files later.
Looking at the data
list
We can use the list command to take a look at the data:
. list
+-------------------------------------------------------------------+
1. | state | region | pop | poplt5 | pop5_17 |
| Alabama | South | 3,893,888 | 296,412 | 865,836 |
|-------------------------------------------------------------------|
| pop18p | pop65p | popurban | medage | death | marriage |
| 2,731,640 | 440,015 | 2,337,713 | 29.30 | 35,305 | 49,018 |
|-------------------------------------------------------------------|
| divorce |
| 26,745 |
+-------------------------------------------------------------------+
+-------------------------------------------------------------------+
2. | state | region | pop | poplt5 | pop5_17 |
| Alaska | West | 401,851 | 38,949 | 91,796 |
|-------------------------------------------------------------------|
| pop18p | pop65p | popurban | medage | death | marriage |
| 271,106 | 11,547 | 258,567 | 26.10 | 1,604 | 5,361 |
|-------------------------------------------------------------------|
| divorce |
| 3,517 |
+-------------------------------------------------------------------+
...and so on.describe
As should be obvious, this usually gives too much information back. A better place to start with a well-formatted data file is to use the describe command:
. describe
Contains data from ../data/census.dta
obs: 50 1980 Census data by state
vars: 12 28 Aug 2012 11:20
size: 2,800
-------------------------------------------------------------------------------
storage display value
variable name type format label variable label
-------------------------------------------------------------------------------
state str14 %-14s State
region int %-8.0g cenreg Census region
pop long %12.0gc Population
poplt5 long %12.0gc Pop, < 5 year
pop5_17 long %12.0gc Pop, 5 to 17 years
pop18p long %12.0gc Pop, 18 and older
pop65p long %12.0gc Pop, 65 and older
popurban long %12.0gc Urban population
medage float %9.2f Median age
death long %12.0gc Number of deaths
marriage long %12.0gc Number of marriages
divorce long %12.0gc Number of divorces
-------------------------------------------------------------------------------
Sorted by: codebook
To get more information about a single variable, thecodebook command is a good option:
. codebook pop
-------------------------------------------------------------------------------
pop Population
-------------------------------------------------------------------------------
type: numeric (long)
range: [401851,23667902] units: 1
unique values: 50 missing .: 0/50
mean: 4.5e+06
std. dev: 4.7e+06
percentiles: 10% 25% 50% 75% 90%
671743 1.1e+06 3.1e+06 5.5e+06 1.1e+07list with if statement
If I add the condition if _n < 11, I can see data for only the first ten states. _n represents the row number of each observation. Since the states are in alphabetical order in the dataset, I can use the logical statement _n < 11 to get the first ten:
. list if _n < 11
+------------------------------------------------------------------------+
1. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Alabama | South | 3,893,888 | 296,412 | 865,836 | 2,731,640 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 440,015 | 2,337,713 | 29.30 | 35,305 | 49,018 | 26,745 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
2. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Alaska | West | 401,851 | 38,949 | 91,796 | 271,106 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 11,547 | 258,567 | 26.10 | 1,604 | 5,361 | 3,517 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
3. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Arizona | West | 2,718,215 | 213,883 | 577,604 | 1,926,728 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 307,362 | 2,278,728 | 29.20 | 21,226 | 30,223 | 19,908 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
4. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Arkansas | South | 2,286,435 | 175,592 | 495,782 | 1,615,061 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 312,477 | 1,179,556 | 30.60 | 22,676 | 26,513 | 15,882 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
5. | state | region | pop | poplt5 | pop5_17 | pop18p |
| California | West | 23,667,902 | 1,708,400 | 4,680,558 | 17,278,944 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 2,414,250 | 21,607,606 | 29.90 | 186,428 | 210,864 | 133,541 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
6. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Colorado | West | 2,889,964 | 216,495 | 592,318 | 2,081,151 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 247,325 | 2,329,869 | 28.60 | 18,925 | 34,917 | 18,571 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
7. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Connecticut | NE | 3,107,576 | 185,188 | 637,731 | 2,284,657 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 364,864 | 2,449,774 | 32.00 | 26,005 | 26,048 | 13,488 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
8. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Delaware | South | 594,338 | 41,151 | 125,444 | 427,743 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 59,179 | 419,819 | 29.80 | 5,123 | 4,437 | 2,313 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
9. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Florida | South | 9,746,324 | 570,224 | 1,789,412 | 7,386,688 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 1,687,573 | 8,212,385 | 34.70 | 104,190 | 108,344 | 71,579 |
+------------------------------------------------------------------------+
+------------------------------------------------------------------------+
10. | state | region | pop | poplt5 | pop5_17 | pop18p |
| Georgia | South | 5,463,105 | 414,935 | 1,231,195 | 3,816,975 |
|-----------------------------------------------+------------------------|
| pop65p | popurban | medage | death | marriage | divorce |
| 516,731 | 3,409,081 | 28.70 | 44,230 | 70,638 | 34,743 |
+------------------------------------------------------------------------+Most of the time, I only want to see a couple of variables. In this case, I’ll use list with what Stata calls a varlist and is in fact just a list of variables. In this case, I only choose state and pop.
. li state pop if _n < 11
+--------------------------+
| state pop |
|--------------------------|
1. | Alabama 3,893,888 |
2. | Alaska 401,851 |
3. | Arizona 2,718,215 |
4. | Arkansas 2,286,435 |
5. | California 23,667,902 |
|--------------------------|
6. | Colorado 2,889,964 |
7. | Connecticut 3,107,576 |
8. | Delaware 594,338 |
9. | Florida 9,746,324 |
10. | Georgia 5,463,105 |
+--------------------------+QUICK EXERCISE
Take a look at deaths in the first 10 states. Which is highest, which is lowest?
Recoding variables
To start off with, I’m interested in knowing which states have the largest proportion of the population under 5. The data only give the total number of people under 5, so I’m going to need a new variable, which will be total population under 5 divided by total population. To create this variable I’ll need Stata’s generate command:
. generate poplt5_pr = poplt5 / popNB: Stata will not allow you to generate a new variable with an old variable’s name. generate poplt5 = poplt5 / pop will not work because you already have a poplt5 variable. This is a feature to make sure you don’t overwrite your data accidentally.
Summarizing data
Now that I have my new variable, let’s use the summarize command to take a look at it:
. summarize poplt5_pr
Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
poplt5_pr | 50 .075981 .0119612 .0585066 .1300186This is nice, but if I’d like even more information I should use the detail subcommand, like so:
. sum poplt5_pr, detail
poplt5_pr
-------------------------------------------------------------
Percentiles Smallest
1% .0585066 .0585066
5% .0595924 .0587786
10% .0629542 .0595924 Obs 50
25% .0698113 .0598551 Sum of Wgt. 50
50% .0750633 Mean .075981
Largest Std. Dev. .0119612
75% .0786851 .0955049
90% .0870086 .096924 Variance .0001431
95% .096924 .0990863 Skewness 1.96293
99% .1300186 .1300186 Kurtosis 9.799138QUICK EXERCISE
Create a variable for the proportion of the population living in urban areas. Use
summarizeto describe your new variable. What’s the mean and median of your new variable?
Using the by and bysort commands
Many times we’d like to summarize a variable by subgroups in the data. For instance, what if we’d like to know which regions have the highest proportions of children under 5? We could try to use the by command like this, by region: sum poplt5_pr, but it won’t work. Stata will refuse to run it because the data are not sorted on the region variable. However, the bysort command gives us an easy way around that problem:
QUICK EXERCISE
Create a table of proportion urban by region. Which region has the highest proportion of people living in cities?
Univariate graphics
histogram
To describe a data point, we can use the histogram command. If we want to save the plot, we use the graph export command:
. histogram poplt5_pr, name(h_poplt5_pr)
(bin=7, start=.05850657, width=.01021601)
. graph export "../plots/h_poplt5_pr.eps", name(h_poplt5_pr) replace
(file ../plots/h_poplt5_pr.eps written in EPS format)
histogram with by
You can combine the histogram command with a by command to show the distribution of a variable by groups:
. histogram poplt5_pr, by(region) name(h_poplt5_pr_reg)
. graph export "../plots/h_poplt5_pr_reg.eps", name(h_poplt5_pr_reg) replace
(file ../plots/h_poplt5_pr_reg.eps written in EPS format)
kdensity
You can also use the kdensity command to describe the data using a kernel density plot:
. kdensity poplt5_pr, name(kd_poplt5_pr)
. graph export "../plots/kd_poplt5_pr.eps", name(kd_poplt5_pr) replace
(file ../plots/kd_poplt5_pr.eps written in EPS format)
QUICK EXERCISE
List state name and population less than 5 if population less than 5 is greater than .1
Bivariate graphics
scatterplot
A scatterplot is a very useful tool for the looking at the relationship between two (or more) variables. Right now I’d like to look at the relationship between the number of children under 5 and the number of people over 65. The variable pop65p is not a proportion, so I need to generate a new proportion variable to get them both on the same scale:
. gen pop65p_pr = pop65p / popWith my new variable, I can now create a scatterplot:
. gen pop65p_pr = pop65p / pop
. // scatterplot of young population as a function of older population
. graph twoway scatter poplt5_pr pop65p_pr, name(sc_poplt5_pr)
. graph export "../plots/sc_poplt5_pr.eps", name(sc_poplt5_pr) replace
(file ../plots/sc_poplt5_pr.eps written in EPS format)
We can add state labels:
. graph twoway scatter poplt5_pr pop65p_pr, ///
> msymbol(none) mlabel(state) name(sc_poplt5_pr_1)
. graph export "../plots/sc_poplt5_pr_1.eps", name(sc_poplt5_pr_1) replace
(file ../plots/sc_poplt5_pr_1.eps written in EPS format)
The labels are too big. We can make them smaller.
. graph twoway scatter poplt5_pr pop65p_pr, ///
> msymbol(none) mlabel(state) mlabsize (tiny) name(sc_poplt5_pr_2)
. graph export "../plots/sc_poplt5_pr_2.eps", name(sc_poplt5_pr_2) replace
(file ../plots/sc_poplt5_pr_2.eps written in EPS format)
EXERCISES
- Create variables for rate of marriages and divorces
- Which region has the highest rates of marriage and divorce in the population?
- What do the distributions of these two variables look like?
- What does a scatterplot say about the possible relationship between the two?
Init: 07 June 2015; Updated: 16 August 2015