Introduction to STATA n n About STATA Basic

Introduction to STATA n n About STATA Basic Operations Regression Analysis Panel Data Analysis

About n STATA is modern and general command driven package for statistical analyses, data management and graphics. n STATA provides commands to analyze panel data (cross-sectional time -series, longitudinal, repeated-measures, and correlated data), crosssectional data, time-series data, survival-time data, cohort study, … n STATA is user friendly. n STATA has an extraordinary set of reference books. n STATA has internet capabilities (installing new features, updating)

Getting ready n n Download statadata. zip from Econ 511 website Unzip file statadata. zip to U: stata

Basic Operations n Entering Data n Exploring Data n Modifying Data n Managing Data n Analyzing Data

Entering Data n Insheet: Read ASCII (text) data created by a spreadsheet (. csv files only) n Infile: Read unformatted ASCII (text) data (space delimited files) n Input: Enter data from keyboard n Describe: Describe contents of data in memory or on disk n Compress: Compress data in memory n Save: Store the dataset currently in memory on disk in Stata data format n Count: Show the number of observations n List: List values of variables n Clear: Clear the entire dataset and everything else n Memory: Display a report on memory usage n Set memory: Set the size of memory

Example n cd u: stata n dir n insheet using hs 0. csv (If file has variable name on the first line) n Save hs n insheet gender id race ses schtyp prgtype read write math science socst using hs 0_noname. csv, clear(If file doesn’t have variable name on the first line) n Count n Describe n Compress n Clear n use hs, clear (only for files in Stata files, can be use over internet) n Memory n set memory 5 m (maximum: 256 MB)

Exploring data n Describe: Describe a dataset n List the contents of a dataset n Codebook: Detailed contents of a dataset n Log: Create a log file n Summarize: Descriptive statistics n Tabstat: Table of descriptive statistics n Table: Create a table of statistics n Stem: Stem-and-leaf plot n Graph: High resolution graphs n Kdensity: Kernal density plot n Sort: Sort observations in a dataset n Histogram: Histogram for continuous and categorical variables n Tabulate: One- and two-way frequency tables n Correlate: Correlations n Pwcorr: Pairwise correlations n Type: Display an ASCII file

Example n n n n n use hs 0, clear Describe List list gender-read Codebook log using unit 1, text replace (open a existing log file called unit 1 which will save all of the commands and the output in a text file and delete the contents and places the current log into the file summarize read math science write display 9. 48^2 (note: variance is the sd (9. 48) squared) summarize write detail sum write if read>=60 sum write if prgtype=="academic“ sum write in 1/40 tabulate prgtype, summarize(read) stem write graph box write log close (close the log file) type unit 1. log (see what is in the log file)

Modifying Data n label data: Apply a label to a data set n Order: Order the variables in a data set n n n n n label variable: Apply a label to a variable label define: Define a set of a labels for the levels of a categorical variable label values: Apply value labels to a variable List: Lists the observations Rename: Rename a variable Recode: Recode the values of a variable Notes: Apply notes to the data file Generate: Creates a new variable Replace: Replaces one value with another value Egen: Extended generate - has special functions that can be used when creating a new variable

Example n n n n Use hs 0 Order id gender label variable schtyp "The type of school the student attended. " label define scl 1 public 2 private label values schtyp scl codebook schtyp list schtyp in 1/10, nolabel encode prgtype, gen(prog) (create a new numeric version of the string variable prgtype) label variable prog "The type of program in which the student was enrolled. " codebook prog list prog in 1/10, nolabel

Example (cont) n rename gender female (easier to work with since we don’t have to deal with 0 s and 1 s) n label variable female "The gender of the student. " n label define fm 1 female 0 male n label values female fm n codebook female n list female in 1/10, nolabel n Gen total = read +write + math n replace total = read + write + socst n label variable total "The total of the read, write and socst. " n list race if race == 5 n recode race 5 =. n list race if race ==. n generate total = read + write + math n sum total n Codebook total n notes race: values of race coded as 5 were recoded to be missing n egen zread = std(read) (using special function std(. )) n save hs 1

Managing Data n Pwd: Show current directory (pwd=print working directory) n dir or ls: Show files in current directory n cd Change directory n keep if: Keep observations if condition is met n Keep: Keep variables (dropping others) n Drop: Drop variables (keeping others) n append using: Append a data file to current file n Merge: Merge a data file with current file

Example We take the hs 1 data file and make a separate folder called honors and store a copy of our data which just has the students with reading scores of 60 or higher n n n n use hs 1, clear Pwd Dir Ls cd honors keep if read >= 60 Describe summarize read save hsgoodread, replace use hsgoodread, clear drop ses save hsdropped, replace describe list in 1/20

Analyzing Data n n n n Ttest: t-test Regress: Regression Predict: Predicts after model estimation Kdensity: Kernel density estimates and graphs Pnorm: Graphs a standardized normal plot Qnorm: Graphs a quantile plot Rvfplot: Graphs a residual versus fitted plot Rvpplot: Graphs a residual versus individual predictor plot Xi: Creates dummy variables during model estimation Test: Test linear hypotheses after model estimation Oneway: One-way analysis of variance Anova: Analysis of variance Logistic: Logistic regression Logit: Logistic regression

Example n use hs 1, clear n ttest write = 50 (This is the one-sample t-test, testing whether the sample of writing scores was drawn from a population with a mean of 50 ) n ttest write = read (This is the paired t-test, testing whether or not the mean of write equals the mean of read) n ttest write, by(female) (This is the two-sample independent t-test with pooled (equal) variances) n ttest write, by(female) unequal (This is the two-sample independent t-test with separate (unequal) variances) n oneway write prog n anova write prog (Both of these commands perform a one-way analysis of variance (ANOVA) n anova write prog female prog*female (the anova command is used to perform a two- way analysis of variance (ANOVA). ) n anova write prog female prog*female read, cont(read) (the anova command performs an analysis of covariance (ANCOVA))

Example (cont) n regress write read female (Plain vanilla OLS regression) n regress write read female, robust (we run the regression with robust n n n standard errors. This is very useful when there is heterogeneity of variance. This option does not affect the estimates of the regression coefficients. ) predict p (The predict command calculates predictions, residuals, influence statistics, and the like after an estimation command. The default shown here is to calculate the predicted scores) predict r, resid (When using the resid option the predict command calculates the residual) pnorm r ( produces a normal probability plot and it is another method of testing whether the residuals from the regression are normally distributed) Rvfplot (generates a plot of the residual versus the fitted values; it is used after regress or anova) rvpplot read (produces a plot of the residual versus a specified predictor and it is also used after regress or anova.

Example (cont) n n n n xi: regress write read i. prog (The xi prefix is used to dummy code categorical variables such as prog. The predictor prog has three levels and requires two dummy-coded variables) test _Iprog_2 _Iprog_3 (The test command is used to test the collective effect of the two dummy-coded variables; in other words, it tests the main effect of prog) xi: regress write i. prog*read (create dummy variables for prog and for the interaction of prog and read) test _Ipro. Xread_2 _Ipro. Xread_3 (tests the overall interaction) test _Iprog_2 _Iprog_3 (tests the main effect of prog) gen honcomp = write >= 60 (create a dichotomous variable called honcomp (honors composition) to use as our dependent variable) tab honcomp The logistic command defaults to producing the output in odds ratios but can display the coefficients if the coef option is used. The exact same results can be obtained by using the logit command, which produces coefficients as the default but will display the odds ratio if the or option is used: n n logit honcomp read female, or

Logistic Regression Classical Regression vs Logistic Regression n All of the previous regression examples have used continuous dependent variables. n Logistic regression is used when the dependent variable is binary or dichotomous. Different Assumptions n The population means of the dependent variables at each level of the independent variable are not on a straight line, i. e. , no linearity. n The variance of the errors are not constant, i. e. , no homogeneity of variance. n The errors are not normally distributed, i. e. , no normaility. Logistic Regression Assumptions: n The model is correctly specified, i. e. , 1. 2. 3. 4. n n the true conditional probabilities are a logistic function of the indpendent variables, no important variables are omitted, no extraneous variables are included, and the independent variables are measured without error. The cases are independent. The independent variables are not linear combinations of each other. Perfect multicolinearity makes estimation impossible, while strong multicolinearity makes estimates imprecise.

Logistic Regression - 2 Logit: n Use admission into a graduate program in which 70% of the males and 30% of the females are admitted n Let P equal the probability of being admitted. n Let Q = 1 - P equal the probability of not being admitted. n Let the odds of a male admitted be odds(M) = P/Q = P/1 -P =. 7/. 3 = 2. 3333 n Let the odds of a female admitted be odds(F) = P/Q = P/1 -P =. 3/. 7 =. 42857 n Let the odds ration, OR = odds(M)/odds(F) = 2. 3333/. 42857 = 5. 44 n The odds if being admitted to the program are about 5. 44 times greater for males then for females. n Let logit(P) = log(odds) = ln(P/Q) = ln (P/1 - P) n This results in the logistic regression equation logit(P) = a + b. X. n In effect, this represents a transformation of the dependent variable such that the resulting logistic regression equation better meets the assumptions of linearity, normality and homogeneity of variance Interpreting logit coefficients: n Logistic slope coefficients can be interpreted as the effect of a unit of change in the X variable on the predicted logits with the other variables in the model held constant. That is, how a one unit change in X effects the log of the odds when the other variables in the model held constant. Interpreting Odds Ratios: n Odds ratios in logistic regression can be interpreted as the effect of a one unit of change in X in the predicted odds ratio with the other variables in the model held constant

Logistic Regression – 3 Sample data set: n n n n n n input apt gender admit 811 710 511 310 511 711 811 511 400 701 301 200 400 200 300 401 300 200 end

Logistic Regression – 4 Example 1: Categorical Independent Variable n n logit admit gender logistic admit gender Example 2: Continuous Independent Variable n n logit admit apt logistic admit apt Example 3: Categorical & Continuous Independent Variables n n logit admit gender apt logistic admit gender apt Example 4: Honors Composition using HSB Dataset n n n n Use hsb 2, clear generate honors = (write>=60) (create dichotomous response variable) tabulate ses, generate(ses) (create dummy coding for ses) logit honors female ses 1 ses 2 read math test ses 1 ses 2 logistic honors female ses 1 ses 2 read math lfit (goodness-of-fit test) lstat

Do file n n n n n Do-files are created with the do-file editor or any other text editor. Any command which can be executed from the command line can be placed in a do-file To open a do file editor: Window – Do-file Editor or Ctrl + 8 set more off use hsb 2, clear generate lang = read + write label variable lang "language score" tabulate lang female tabulate lang prog tabulate lang schtyp summarize lang, detail table female, contents(n lang mean lang sd lang) table prog, contents(n lang mean lang sd lang) table ses, contents(n lang mean lang sd lang) correlate lang math science socst regress lang math science female set more on

Do file – cont. Look at the commands in a do-file that contains: n . type hsbbatch. do To run the do-file. n do hsbbatch n From do file, choose Tools - Do

Panel Data Creat the do file as followed n n n n n set matsize 160 use http: //www. ats. ucla. edu/stata/stat 130/depress, clear sort group by group: summarize pre dep 1 dep 2 dep 3 dep 4 dep 5 dep 6 corr pre dep 1 dep 2 dep 3 dep 4 dep 5 dep 6 graph dep 1 dep 2 dep 3 dep 4 dep 5 dep 6, matrix half ttest pre, by(group) /* check to see if the groups differ on the pretest depression score */ hotel dep 1 dep 2 dep 3 dep 4 dep 5 dep 6, by(group)/*There isn't much of a difference between groups on the pretest so let's try a Hotelling's T 2 Using Hotelling's T 2 we find a significant difference between the two groups. The T 2 did not make use of any of the information concerning the pretest but that's okay for the moment especially since we know that the pretest differences were not significant. */ reshape long dep, i(subj) j(visit) regress dep pre group visit glm dep pre group visit, fam(gaus) link(iden) xtgee dep pre group visit, fam(gaus) link(iden) i(subj) t(visit) corr(ind) /*The three previous analyses provide identical incorrect results. The common thread among them is that they all assume that the observations within the subjects are independent. This seems, on the face of it, to be highly unlikely. Scores on the depression scale are not likely to be independent from one visit to the next. Of the three, only xtgee makes the assumption concerning the correlations explicit. */ xtcorr /* The xtcorr command shows structure of the correlation matrix*/ /* xt commands are used with cross-sectional time-series data */ xtsum dep

Panel data 2 n n n n /*We can analyze these data using compound symmetry for the correlational structure. This approach can be tried using exchangable for the correlation matrix in xtgee */ xtgee dep pre group visit, fam(gaus) link(iden) i(subj) t(visit) corr(exc) xtcorr /*Note in particular the change in the standard errors between this analysis and the previous one. Now let's try a different correlation structure, auto regressive with lag one. */ xtgee dep pre group visit, fam(gaus) link(iden) i(subj) t(visit) corr(ar 1) /*back up and reconsider the group by visit interaction. We will try a model with the interaction using the ar 1 correlations. */ generate gxv = group*visit xtgee dep pre group visit gxv, fam(gaus) link(iden) i(subj) t(visit) corr(ar 1) /* The group by visit interaction still is not significant even though this may be a better approach for testing it. So far we have been treating visit as a continuous variable. Is it possible that our analysis might change if we were to treat visit as a categorical variable, the way that the anova did? Let's try one last analysis using xi to create dummy variables on-the-fly. */ xi: xtgee dep pre group i. visit, fam(gaus) link(iden) i(subj) corr(ar 1)

Searching for help n n n n The help command can be used from the command line or from the Help window. To use help the command must be spelled correctly and the full name of the command must be used. help contents will list all commands that can be accessed using help if help anova help regress The search command searches for information in Stata manuals, FAQs, and Stata Technical Bulletins (STBs). The search options include: manual which restricts searches to the Stata Manual; author when searching for an author by name; stb which restricts searhes to STBs; faq which restricts searches to FAQs. The search command can be used from either the command line or the Help window. search if search regression search ttest, manual Each copy of Stata comes with a built-in tutorital. Typing tutorial brings up information about the tutorials. tutorial regress will bring up the tutorial on regression. tutorial regress

End of Session