Agriculture Reference
In-Depth Information
tabulate
tractor repair
[fweight =
number
], chi2 expected
restore
he
tabulate
command entered with (
fweight
=
number
) uses
the
number
variable as a frequency weight to calculate the out-
put, which is exactly the same as shown above. The
preserve
and
restore
commands preserve and restore the original dataset after
using the
contract
command.
In previous chapters, the data generally had to meet certain under-
lying criteria, such as normality, additivity, homogeneous variances,
etc. In some cases, it was possible to transform data to meet these cri-
teria. There are also tests that do not require these underlying assump-
tions. These methods are often referred to as nonparametric tests. In
general, if the underlying assumptions for the parametric tests are
true, then these nonparametric tests are not as powerful.
One such test is called the Sign Test. The Sign Test is much like the
paired t-test without any underlying assumptions about the popula-
tions. This test is an evaluation of medians to see if two medians dif-
fer significantly from 0. In general, this test works better with 20 or
more paired data points. The advantage of not having specific require-
ments in the population is offset by the loss of information concerning
the magnitude of the differences. Open the dataset Sign Test Food
Products.dta. This is a dataset of 22 people rating two different snacks
on a scale of 1-20 with 20 considered the best (Davis, 2000, p. 198).
There is an error in the dataset as listed in Davis' text, so the results
will be different here. Enter the following command:
signtest
tomato = apricot
This results in the following output:
Sign test
sign | observed expected
-------------+------------------------
positive | 7 11
negative | 15 11
zero | 0 0
-------------+------------------------
all | 22 22
Search WWH ::
Custom Search