Stata Seminar

1
Stata Seminar

• Session 1
• Francisco Jose Gonzalez Carreras
• fjg23_at_sussex.ac.uk

2
Source

• This is the source used. The sessions will be
• First session Hands on session. (Chapter 1)
• Second session Grammar of Stata. (Chapter 3)
• Third session. Creating and changing variables.
  (Chapter 5)
• Fourth session. Charts and Linear Regression.
  (Chapters 6 and 8).
• This is the 2005 edition, there is a new one
  forthcoming.
• Not in library but could be borrowed by
  interlibrary loan (need to pay 2, though)

3
Starting Stata

• Download the data from
• http//www.stata-press.com/data/kk.html kk.zip
  file
• To start the session Start gt All programs gt
  Intercooled Stata.

4
Screen
Pop-Up menu
Past commands appear here
Results appear here
Working directory displayed here
Variable list displayed here
Commands typed appear here
5
Stata Screen

• Change the default windows
• Right click the mouse in the results screen and
  you can change the font of the results windows.
• You can also move windows around
• If something was changed and you want to restore
  the original settings, Pop Up menu gt Prefs gt
  Manage Preferences gt Load preferences gt Factory
  settings (In version 10 you have to go Pop Up
  menu gt Edit gt Preferencesgt and the rest is the
  same)

6
Analysis Input commands

• Type d in the command window and press Return
• d is the abbreviation of describe, a command that
  describes file.
• The number of observations and variables is zero
  we have not loaded any file.
• Memory is of working memory being used. Data
  are loaded in the RAM memory.
• Sorted by no sorting criteria.
• We have not loaded any data, lets load a file.

7
Analysis Directory

• Command cd change directory. We have to move to
  the directory where the files are. Type cd
  c\data\kk to move. This allow us to name the
  file without having to write the whole path of
  the file.
• Dir will show all the files containded in the
  file.
• See that more- appears click enter and you
  advance a file, click space bar and you will see
  the next screen. Click q to stop results coming.
• You can use dir .dta, which will show you all
  the files with dta extension.

8
Analysing data loading data

• To use a file, type
• use data1
• This command loads stata file into the working
  memory (RAM).
• Default memory size is 1 mega and sometimes you
  will need to set a bigger capacity. For big files
  set memory would be needed
• Stata assumes it is a .dta file.
• Then type describe

9
Analysing data Variables Observations
10
Analysing data Variables Observations

• The file is a subsample of the German
  Socioeconomic Panel (GSOEP). It is a survey
  taking place since 1984 in which the same
  households, families, individuals are interviewed
  once a year.
• In the screen
• Observations are 3,340 and the nr of variables
  47. This means that 47 pieces of information are
  stored for each individual.
• The first, persnr is the first variable and does
  contain only the code that is unique for each
  individual. Sometimes you will need to create
  this from other information.
• Storage type has to do with the size and is
  important to save resources (more on this next
  sessions)
• Labels is a brief description of the variable
  (more coming)

11
Analysing data Looking at data

• We have too many observations, so we get rid of
  some. Type
• drop ymove-np9507 (get rid of the range of the
  variables included in the command, keep gets rid
  of the variables not included)
• To have a look all observations type list (Then
  q to stop more screens!!)
• Too much information, not operational. We will
  reduce it.
• Lets focus on the second Man, was born in 1971,
  household head, single..
• Missing value . was not questioned of the
  person did not respond.

missing value
12
Analysing data Looking more carefully

• Listing data in this fashion is not useful so we
  will be more specific.
• We could type list to list just a number of
  variables. Type list gender income
• but again we have more than 3,000
  observations!!.
• To narrow down our look we will use first the in
  qualifier. This qualifier limits by the position
  of the observation in a particular order. Type
• sort income
• list gender income in 1/10

13
Analysing data looking more carefully

• What does these commands do????
• sort income sort the data in ascending order, so
  the person with the lowest income is the first.
  This establish the order.
• list gender income in 1/10 list the first 10
  observations. It will show gender and income
  corresponding to the ten observations
  (respondents) with the lowest income (remember
  we sorted in ascending order by income)
• What would this do?
• list gender income in 2/4
• Individuals from the second to the fourth

14
Analysing data Summary statistics

• To obtain summary statistics about income, type
• Summarize income
• The information about is the nr. of observations
  used to calculate the arithmetic mean, the
  standard deviation, the minimum and the maximum.
  You have only 3,034 observations because some of
  them were set to missing (.) and they are not
  taken into account when doing the calculations
• You can summarize a list of variables simply by
  adding more in the list. If you want to summarize
  all the variables, just type summarize (also sum
  as an abbreviation would work)

15
Analysing data if qualifier

• What if we wanted to have a look at the
  difference in income between women and men? We
  use the if qualifier and summarize data
  conditional on the variable meeting the if
  condition. Type
• summarize income if gender1
• summarize income if gender2
• The first summarize only the observations in
  which gender is equal to a particular value. 1
  refers to males and 2 to females in this survey.
  See the difference in income.
• The double equal is necessary, otherwise it will
  show invalid syntax

16
Analysing data missing values

• Men seem to earn more than women. But these
  averages are calculated taking into account those
  observations with income0. These might be to be
  more frequent among women so in order to compare
  only individuals with positive income we can
  either type
• sum income if gender1 incomegt0
• sum income if gender2 incomegt0
• or recode 0 incomes to . (missing) so that
  they will not be taken into account when
  calculating the average. Type
• mvdecode income, mv(0.a)
• sum income if gender1
• sum income if gender2

17
Analysing data by prefix

• A prefix is a command that is written in front of
  the actual stata command.
• It has two parts
• prefix itself, by
• variable list, in our example only gender.
• Structure would be
• prefix command actual command
• In the case of by, the actual command is repeated
  for all the categories in the prefix list or
  bylist.
• A condition is that the data have to be sorted by
  the variables in the bylist
• Type
• sort gender
• by gender summarize income

18
Analysing data missing recoding by prefix
Same mean
19
Analysing data Command options

• Options are command specific, unlike in and if
  qualifiers or by prefix.
• They are written after the actual command,
  following a comma.
• In the case of summarize, the detail option
  will give much more information about the income
  distribution skewness, kurtosis
• Type
• sum income, detail

Median
Moments
20
Analysing data Frequency tables

• The command that generates frequency tables is
  tabulate (or tab), which has to be followed by
  one or two variables, generating one way
  frequency table or two way frequency table.
• Type
• tabulate gender
• tabulate emp gender
• First variable is the row variable, second
  variable is the column variable.
• Options for this command are row or column which
  return the row and column percentages

21
Variable labels and value labels

• See the differences between the first and the
  second table. In the second we only have the
  values that correspond to the different types of
  employment status.
• label of the variable is a brief description of
  the variable. Lets change it typing (does not
  matter if it already had one)
• label variable emp Status employment in 97
• label values is the label for the different
  values. In income, the label for value 1 was
  male and the label for value 2 was female.
  This variable has seven different values. Lets
  label the values. Type (not breaking the line)
• label define emplb 1 Full time 2 Part time 3
  Retraining gt 4 irregular 5 not working 6
  military service 7 gt gtunemployed , modify
• label values emp emplb
• Labels are stored in emplb. They can be assigned
  to any other variable with same values. Let
  tabulate again to see the changes
• tab emp gender, column nofreq

22
Variable labels and value labels
was Employment Status 1997
value labels created
23
Analysing data Graphs

• Part-time employment or unemployment is more
  frequent among women. Maybe income differences
  are due to employment status.
• Graph the distribution might give us a quick and
  very informative look over the data. Type
• graph box income, over(emp)
• To get a box-and-whisker plot, this result in a
  graph with one distribution graph over each group
  of emp ,(over (emp))
• Outliers are the dots. Income are skewed for all
  subgroups. Median for full time is higher than
  for the rest. If there are relatively more part
  time women represented, we might think that
  income inequality could be due to division of
  labor within the couples than to gender
  discrimination. We first must control for
  employment status

24
Analysing data graphs
Outliers
Third quartile
Median
First quartile
employment status
25
Getting help

• How to find out about the effects of gender and
  employment status on income?. Regression
  analysis. How to do it? Lets have a look at the
  help
• command search looks in all stata resources some
  topic that might be linked to your search. Type
• search Linear Regression
• search model
• search OLS
• Also you can use the help command to get
  information about a command (now that we know
  that we should use regress). Type
• help regress
• You find the syntax, explanation and description
  of the available options (options, as formerly
  said are command specific)
• Pop up menu Help gt Search or Help gt Stata
  command

26
Getting Help
27
Analysing data Recoding variables

• We have the dependent variable, income, and two
  independent variables gender and employment
  status.
• Gender dichotomous variable. They conventionally
  take the values of 0 or one in regressions. We
  recode the variable to make men 1, women 0.
  Type
• generate men1 if gender1
• replace gender0 in gender2
• Create a variable 1 if gender1 and missing
  otherwise. Then replace with 0 those missing
  values that meet the criteria of gender2
• With employment status we will do something
  similar. This variable is not dichotomous. We
  will do just
• generate fulltime1 if emp1
• replace fulltime 0 if emp2
• because the analysis will be limited to full
  time/part time.

28
Analysing data Linear Regression

• We will run a basic linear regression with the
  data at hand. We saw that we needed to use
  regress. We type regress followed by the
  dependent variable followed by the independent
  variables. Type
• regress income men fulltime
• Interpretation
• average monthly income for individuals with
  income0 and fulltime0 (part time women
  employees) is 965.
• female full time workers earn on average 806
  more.
• independent of full time/part time, men earn on
  average 451 more than women.
• Therefore, income inequality cannot be explained
  by the higher proportion of female part time
  workers in the data file.

29
Analysing data Linear Regression
30
Do files

• To reproduce the results in your session do
  files.
• Text file where you save your commands in order
  to store your work sessions.
• Type
• doedit
• Opens the do file. The first line establish the
  version so that the do file can be run with any
  future version. We have written the commands that
  were necessary to do the regression as above.
• Once you have copied the commands File gt Save as
  gt an1.do in the current directory. Next type
• do an1.do
• You can run all the commands again!!!!

31
Do files
32
Exiting Stata

• Once you have saved your session or your work in
  your do file, it is better to leave stata without
  saving changes.
• Changes in do files are easy do to, changes in
  the original database might not be possible to
  undo. HAVE ALWAYS MORE THAN ONE COPY OF THE
  ORIGINAL DATABASEjust in case.
• If you want to save changes, save them in a new
  file, typing, for instance
• save mydata
• Then you can exit Stata exit, clear