Agriculture Reference
In-Depth Information
Stata offers a number of commands that can be used to determine
normality of the residuals. This includes the
rvpplot
and
rvf-
plot
commands that were shown in Chapter 10, Correlation and
Regression. Data in these scatterplots should appear randomly about
0 on the
y
-axis.
Stata offers additional tests for checking normality including
hettest
,
swilk
,
sfrancia
,
sktest
, and
ladder
. The ladder
command also evaluates several transformations to determine if they
are normally distributed as well. Enter the following command after
the ANOVA has been calculated:
hettest
which results in the following output:
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of plantcount
chi2(1) = 144.96
Prob > chi2 = 0.0000
Using a standard, such as 0.05, for probability; values lower than this
would indicate a nonnormal distribution (Prob>chi
2
). This test indi-
cates that the null hypothesis (H
o
) should be rejected because the chi
2
probability is highly significant. Other commands, such as
swilk
and
sfrancia
, also indicate whether the data are normally distrib-
uted. Enter the following commands and see the output:
swilk
plantcount
Shapiro-Wilk W test for normal data
Variable | Obs W V z Prob>z
-------------+--------------------------------------------------
plantcount | 120 0.70051 28.819 7.530 0.00000
sfrancia
plantcount
Shapiro-Francia W' test for normal data
Variable | Obs W' V' z Prob>z
-------------+--------------------------------------------------
plantcount | 120 0.69665 32.103 6.938 0.00001
Search WWH ::
Custom Search