Agriculture Reference
In-Depth Information
display (
.016989407
+(
2
*
.02406755
))/(
4
*
2
)
The result is 0.00814056.
Next we wish to calculate the CV or standard error of the treat-
ment mean expressed as a percent. This formula is
2
=
100
σ
CV
X
The overall mean (
X
) is required to calculate the CV and can be
acquired with the command
summarize
lwratio
he
summarize
command computes several statistics including the
mean, which is 1.307622. Finally, to calculate the CV, enter
display
100
*sqrt(
.00814056
)/
1.307622
This calculates to 6.8999338%. At this point it is possible to substitute
different numbers of subsamples or replications in the above formulas
to see what effect it has on the CV and, thus, the precision of the
experiment. Increasing either subsamples or replications will lower
the CV. The right combination should have a reasonable CV while
not consuming too many resources. In this case, if the number of sub-
samples is increased to 10, the CV is only reduced from the original
6.9% to 6.1%. Adding one more replication also only reduces the CV
to 6.2%.
Using a somewhat different approach with the above information,
it can be used to estimate the number of subsamples based on the
level of significance required and the margin of error as a fraction of
the treatment mean. The following formula can be used to calculate
this value:
( ()
2
σ
2
Z
a
s
n
=
(
)(
)
−
( ()
2
2
2
2
rD
XZ
σ
a
e
where
Z
a
is the standard normal density,
σ
s
2
is the subsample vari-
ance,
r
is the number of replications,
D
is the margin of error as a
Search WWH ::
Custom Search