Agriculture Reference
In-Depth Information
12
B
inary
, o
rDinal
,
anD
c
ateGorical
D
ata
a
nalySiS
In previous chapters, we have dealt primarily with continuous data,
such as yield. There are occasions where data are not continuous.
For example, some data will have only two possible values, such as
whether a plant is diseased or healthy. Other examples are sex, alive
or dead, etc. It may be useful in such cases to know the probability of
a specific ratio of events. For example, sex ratios between males and
females is approximately 50%. Not every sample from a population is
going to have exactly half male and half female individuals, however.
If you took a sample of 20 individuals, it would not be unusual to have
9 males and 11 females and, although a rarer event, it is even pos-
sible to have all 20 of the individuals be either male or female. Such
binomial events can be calculated. Open the Binomial.dta dataset and
enter the command
bitest
sex ==.5,
detail
This results in the following output:
Variable | N Observed k Expected k Assumed p Observed p
----------+-----------------------------------------------------
sex | 20 4 10 0.50000 0.20000
Pr(k >= 4) = 0.998712 (one-sided test)
Pr(k <= 4) = 0.005909 (one-sided test)
Pr(k <= 4 or k >= 16) = 0.011818 (two-sided test)
Pr(k == 4) = 0.004621 (observed)
Pr(k == 15) = 0.014786
Pr(k == 16) = 0.004621 (opposite extreme)
Datasets of this type can only have data as either 0 or 1, representing
binomial data. In this case, it could be interpreted that 1 is female
and 0 is male. Look at the value Pr(K == 4), which is 0.004621. This
215
Search WWH ::
Custom Search