Page 42 - 2000 (5) Agricultur Census 2000 Balochitan Province

half strata divided by the sub- u n
division estimates as follows :- I ( RVXWs X'*n
)
nn g
n n L' Yn,1 -Yn'z RVXW =
Var.(Y) = tc.V.(Y)7' =Z [- ]'z un
h'n (r X*nf
Y g
Similarly, R.V. for the sub-
division estimate of units having R.V. estimates for provinces and
the bttribute is :- country levels have also been
nn worked out on the same pattern
n n L' Xn'1 Xn,, as at the district level.
-
Var.(X) = [C.V.(X)]'= I [- ]'
h'n (2) Non-Sampling Errors
X
The errors other than those due
Per cent Sampling Error, i.e. to sampling are called non-
sampling errors. These include
(C.V.) = lR.V. x 100 errors in sampling frame,

selection of the sample,
b) Estimates at District, Province enumeration, incorrect response
and Country Levels due to biases and ignorance,
errors in recording answers,
lf the R.V. for the aggregate of processing errors, etc.
characteristic (W) for sub-
division (g) is denoted by Interestingly, sarirpling and non-
RWWg and Y*g is the estimate sampling errors are negatively
for the aggregate (Y) of related. ln other words, an
characteristic (W) for sub- increase in sampling error would
division (g), then the R.V. for the cause a decrease in the non-
district with g = 1 to u sub- sampling error and vice 'versa.
divisions in the district is Normally, oh increase in
sampling error is due to
u n decrease in sample size.
I ( RVYWg Yt*n Therefore, it is conceivable that
)
g the estimate of a given item with
RVWV = very low sampling error might in
u fact be a bad estimate, as it is
(I Y'n f amenable to have substantial

g non-sampling errors. A "good"
and
estimate is, therefore, the one in

XXXVIII