Stat Computing > Stata > Notes

Stata Class Notes
More Statistics

1.0 Stata commands in this unit

. ttest
oneway
anova
test
regress
predict

2.0 Demonstration and Explanation

. use hsb2, clear

We will continue to use the Highschool and Beyond dataset. The file 'hsb2.dta' has correct and complete data for all of the variables.

2.1 t-tests

. ttest write=50

This example involves the single-sample t-test, testing whether the sample was drawn from a population with a mean of 50. By the way, the standardized writing test in this sample was normed nationally with a mean of 50.

. ttest write=read

This example makes use of the t-test for dependent samples. In this case, we are testing whether there is a significant difference betweent the math and the science test scores.

. ttest write, by(female)
ttest write, by(female) unequal
sdtest write, by(female)

The t-test for independent groups comes in two varities: pooled variance and unequal variance. We want to look at the differences in writing test scores between 'school types.' We will begin with the ttest for independent groups with pooled variances and compare the results to the ttest for independent groups using unequal variance.

There is a test for heterogeneity of variance, sdtest, but it is overly sensitive to nonnormality and statisticians do not recommend using it to screen for heterogeneity of variance.

2.2 Analysis of Variance

. oneway write prog, tabulate
anova write prog
sort prog
by prog: summarize write
table prog, contents(n write mean write sd write)

Here are two different ways to perform a one-way analysis of variance (ANOVA). They both give the exact same answer. The most visible difference is that oneway includes a test for homogeneity of variance.

. anova write female prog female*prog

This example demonstrates a 3 X 3 factorial analysis of variance.

2.3 Regression

. regress write read
regress write read, beta
predict pre1
generate pre2 = 23.95944 + .5517051*read
list pre1 pre2
graph pre1 write read, symbol(io) connect(L.)
graph pre1 write read, symbol(io) connect(L.) jitter(2)

These are two examples of simple linear regression. The first one displays confidence intervals for the regression coefficients while the second one displays standardized regression coefficients along with the 'regular' regression coefficients. The predict command computes a predicted science score for each observation. Compare 'pre1' with 'pre2' that was created using the generate command. The graph command, in this example, displays a scatter plot of read and write along with showing the regression regression line of science on math.  The second example uses the jitter option to help see the points where there are multiple observations on one point.

. regress write read math
regress write read math female

This time we have two examples of a multiple regression, the first one with two predictor variables and the second one with three.

2.4 Comments on anova & regress

anova & regress are just two of many estimation procedures available in Stata. A partial list is given in the table below:

Some Estimation Procedures in Stata

anova	analysis of variance and covariance
arch	autoregressive conditional heterosce. family of estimators
arima	autoregressive integrated moving average models
bsqreg	quantile regression with bootstrapped standard errors
clogit	conditional logistic regression
cnreg	censored-normal regression
cnsreg	constrained linear regression
ereg	maximum-likelihood exponential distribution models
glm	generalized linear models
glogit	weighted least squares logit on grouped data
gprobit	weighted least squares probit on grouped data
ivreg	instrumental variable and two-stage least squares regression
lnormal	maximum-likelihood lognormal distribution models
logistic	logistic regression
logit	maximum-likelihood logit regression
mlogit	maximum-likelihood multinomial logit models
mvreg	multivariate regression
nbreg	maximum-likelihood negative binomial regression
nl	nonlinear least squares
ologit	maximum-likelihood ordered logit
oprobit	maximum-likelihood ordered probit
poisson	maximum-likelihood poisson regression
probit	maximum-likelihood probit estimation
qreg	quantile regression
reg3	three-stage least squares regression
regress	linear regression
rreg	robust regression using IRLS
sureg	seemingly unrelated regression
tobit	tobit regression
vwls	variance-weighted least squares regression
zinb	zero-inflated negative binomial model
zip	zero-inflated poisson models

test & predict are commands that can be used in conjuction with estimation procedures. There are too many combinations of estimation, predict and test to get into in this class, other than to say that they provide very powerful tools for researchers and are worth the time spent learning them.

3.0 Try the commands on your own

. use hsb2, clear
ttest math=50
ttest math, by(sch)
oneway math sci
oneway math prog
anova math prog ses prog*ses
test ses / prog*ses
regress science math
regress science math, beta
regress science math read
regress science math read write

4.0 Can you answer these questions?

1. Could this sample have come from a population with a mean reading score of 53?
2. Is there a significant difference in the means of low SES vs middle SES groups?
3. Are the population means for social studies equals for all levels of SES?
4. What is the regression equation for predicating social studies from reading and writing scores?

5.0 For More Information

• Statistics with Stata 5 Chapters 5 & 6
• Stata Learning Modules
  A Statistical Sampler in Stata