In this class we’ll continue exploring some of Stata’s basic functionality.
do filesAll *.do files for this course should follow the outline provided by the example do files. An example template is also posted on the course website. The point is to have a standard structure that will make sense for every new *.do file you create.
*.ado files)One of Stata’s great virtues is that it is highly extensible. Users are frequently creating new commands for Stata and posting them on the web. To get these commands, we use the net search command. From there, a list of possible matches will be presented. After choosing the one you want, go ahead and download it into Stata. You now have access to the new command.
. net search renvars
(contacting http://www.stata.com)
4 packages found (Stata Journal and STB listed first)
-----------------------------------------------------
dm88_1 from http://www.stata-journal.com/software/sj5-4
SJ5-4 dm88_1. Update: Renaming variables, multiply and... / Update:
Renaming variables, multiply and systematically / by Nicholas J. Cox,
Durham University, UK / Jeroen Weesie, Utrecht University, Netherlands /
Support: n.j.cox@durham.ac.uk, j.weesie@fss.uu.nl / After installation,
dm88 from http://www.stata.com/stb/stb60
STB-60 dm88. Renaming variables, multiply and systematically / STB insert
by Nicholas J. Cox, University of Durham, UK / Jeroen Weesie, Utrecht
University, Netherlands / Support: n.j.cox@durham.ac.uk
j.weesie@fss.uu.nl / After installation, see help renvars
cleanchars from http://fmwww.bc.edu/RePEc/bocode/c
'CLEANCHARS': module to replace specific characters or strings in variable
names and/or variable labels and/or string variable values and/or value
label names and levels with stated characters/strings (using 1-1 or m-1
match) / cleanchars is a program that helps out with replacing /
renvarlab from http://fmwww.bc.edu/RePEc/bocode/r
'RENVARLAB': module to rename variables, with option of using variable
labels to create new variable names / This command is an extension of
renvars (also available from / SSC), which renames a list of variables by
applying the given / transformation to all of the variables. It has all ofStata data files end in *.dta. They are easily loaded with the use command. Some datasets, like the one we will use today, can be downloaded from the Stata website directly with the webuse command. While we won’t really use these toy datasets in our research, they can come in handy for small exercises (such as we will do today). They are also the datasets used in the Stata help files.
. webuse school, clearoutsheet datasetStata can export data in almost any commonly-used format. The most commonly used format for data files is in what’s called ASCII delimited form, with a delimiter of either a comma or a tab. To export a dataset we use the outsheet command. You can specify which variables from the current dataset you’d like to export. *.csv is a good format for Microsoft Excel files.
. outsheet using "../data/school_data.csv", comma replaceExport the current dataset in tab delimited format, then go ahead and open up the new dataset in Excel.
insheet datasetSimilarly, Stata can read in data in most any format using the insheet command:
. insheet using "../data/school_data.csv", comma clear
(11 vars, 95 obs)describe datasetOnce you have your dataset in memory, you can describe it to get a quick overview of what you have:
. describe
Contains data
obs: 95
vars: 11
size: 1,900
-------------------------------------------------------------------------------
storage display value
variable name type format label variable label
-------------------------------------------------------------------------------
obs byte %8.0g
pub12 byte %8.0g
pub34 byte %8.0g
pub5 byte %8.0g
private byte %8.0g
years byte %8.0g
school byte %8.0g
loginc float %9.0g
logptax float %9.0g
vote byte %8.0g
logeduc float %9.0g
-------------------------------------------------------------------------------
Sorted by:
Note: Dataset has changed since last saved.Properly labeling datasets and variables will make life much easier for you and anyone else who uses your dataset. To label an entire dataset, use the label data command:
. label data "Voting on school expenditures"To label variables, use the label variable command:
. label variable loginc "Log of income"
. label variable vote "Voted for public school funding"Make sure that your variable labels are highly descriptive and directional (e.g., instead of labeling a binary variable gender label it female). We can see our labels if once again describe our data:
. describe
Contains data
obs: 95 Voting on school expenditures
vars: 11
size: 1,900
-------------------------------------------------------------------------------
storage display value
variable name type format label variable label
-------------------------------------------------------------------------------
obs byte %8.0g
pub12 byte %8.0g
pub34 byte %8.0g
pub5 byte %8.0g
private byte %8.0g
years byte %8.0g
school byte %8.0g
loginc float %9.0g Log of income
logptax float %9.0g
vote byte %8.0g Voted for public school funding
logeduc float %9.0g
-------------------------------------------------------------------------------
Sorted by:
Note: Dataset has changed since last saved.For many binary/categorical variables, you’ll want to say what the underlying levels of the variable in the dataset mean. This is a two-part process. First you define the label values through label define; next you apply them to your particular variable with label values <variable> <label>:
. tab vote
Voted for |
public |
school |
funding | Freq. Percent Cum.
------------+-----------------------------------
0 | 36 37.89 37.89
1 | 59 62.11 100.00
------------+-----------------------------------
Total | 95 100.00
. label define voteopts 0 "no" 1 "yes"
. label values vote voteopts
. tab vote
Voted for |
public |
school |
funding | Freq. Percent Cum.
------------+-----------------------------------
no | 36 37.89 37.89
yes | 59 62.11 100.00
------------+-----------------------------------
Total | 95 100.00Transforming a variable refers to using an operation to create a new version of an existing variable. In our dataset, both income and household spending on education are expressed in terms of the natural log of the existing variable. If we want to have the values of these variables in nominal scale, we need to use the gen command to create a new variable:
. gen inc = exp(loginc)
. sum loginc inc
Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
loginc | 95 9.971017 .4118853 8.294 10.82
inc | 95 23093.31 8871.35 3999.8 50011.07Recoding a variable involves changing the values of a variable based on its existing variables. We often want to recode variables in order to make them more useful for whatever analysis we’re working on. I’ll show you two different ways of recoding, one based on the gen command and a more complicated one using the egen commend with the recode command:
. gen inc_bin = 0
. replace inc_bin = 1 if inc > r(mean)
(30 real changes made)
. egen inc_q = cut(inc), group(4)
. recode inc_q (0 = 1 "First Quartile") ///
> (1 = 2 "2nd Quartile") ///
> (2 = 3 "3rd Quartile") ///
> (3 = 4 "4th Quartile"), gen(new_inc_q)
(95 differences between inc_q and new_inc_q)Create a variable that is equal to 1 if it’s greater than the median of income. Properly label the variable and its values.
Computing a variable involves using the values of other variables to create a new variable. For instance, to calculate an effective property tax rate, we want to divide the property tax value by income:
. gen ptax = exp(logptax)
. gen taxrate = ptax / incInit: 06 June 2015; Updated: 13 August 2015