Please Note: This page has been updated to Stata 12 as of May, 2012.
use https://stats.idre.ucla.edu/stat/data/crf24, clear
* descriptive statistics by group or cell:
table a b, contents(freq mean y sd y) row col
* alternative method
egen grp=group(a b), label
tabstat y, by(grp) stat(n mean sd var)
* graph interaction/plot means by cell:
anova y a b a#b
margins a#b
marginsplot, x(a) // place a on the x-axis
marginsplot, x(b) // place b on the x-axis
* Levene’s test of heterogeneity of variance:
robvar y, by(grp) // W0 is the Levene’s test
* visual check of normality
* histograms
histogram y, by(grp) normal
* kernal density plots
forvalues i = 1/8 {
kdensity y if grp==`i’, normal name(cell`i’, replace)
}
* normal probability plot
forvalues i = 1/8 {
pnorm y if grp==`i’, name(pnorm`i’, replace)
}
* oneway anova:
anova y b
* factorial anova:
anova y a b a#b
anova y a b / a#b / // using interaction as error term
* nested anova:
anova y a b|a // using residual as error term for both a and b|a
anova y a / b|a / // using b|a as error term for a and residual for b|a
* repeated measures anova:
* data are long
use https://stats.idre.ucla.edu/stat/data/spf24, clear
* randomized block design
anova y b s, repeated(b)
mat lis e(Srep) // check variance-covariance matrix
* Tukeys test for additivity in randomized block designs:
anova y b s
predict yhat
generate ystar = (yhat – 5.375)^2 /* 5.375 is the grand mean */
anova y a s c.ystar // if ystar is nonsignificant then model is additive
* split-plot factorial design
anova y a / s|a b a#b/, repeated(b)
* tests of simple main effects:
use https://stats.idre.ucla.edu/stat/data/crf24, clear
anova y a b a#b
contrast a@b
contrast b@a
* pairwise comparisons:
pwcompare b, mcompare(tukey) effects
* trend analysis
contrast p.b
* user defined contrasts:
contrast{b -3 1 1 1} // 1 vs average of 2, 3, and 4
* analysis of covariance:
use https://stats.idre.ucla.edu/stat/stata/examples/kirk/crac4, clear
anova y a c.x
graph slopes
sum x // find min and max
margins a, at(x=(32 89))
marginsplot, x(x)
* adjusted cell means
margins a
* pairwise comparisons of adjusted means
pwcompare a, mcompare(tukey) effects
* check homogeneity of slopes
anova y a c.x a#c.x