Page 36 - 1990(3) Cesus Of Agriculture, provience report-Punjab

Vol . 3 1. Tabulations for North- sampling Erro r
.~p ecial ern Areas including
Reports Northern Area level Sampling error is estimated to
tabulations both in acres . measure the precision/consistency
and hectares and of sample results, i . e. the close-
District level tabulations ness of sample results to those ob-
in acres only. tained through complete count con-
ducted With ~xactly similar ar-
2. Important Aggregates at rangements of training, enumera-
Sub-division level in tion, supervision, processing_, etc ._ -
acres only. If estimates for an item ~re
developed from all possible sampl~s
3 . Mouza stitistics-1988 of same size, drawn with similar
procedure from a population ~ their
In · order · to publish census mean would be - equal to the mean ·
-: tabulations free of typographical based upo·n complete cou_nt and their
errors and to . curtail expenditure dispersion about that _ niean ,
-and save printing time , · the com- measured in terms of vari ance, is
puter _ print - outs of ta-bulations called_ sampling variance. _ Tne
have - been reduced in size and square root ·of sampling variance is
reproduced with the _help of laser · known as sampl i ng error . Further ,
printer and off-set printing it would be noted that estimates
proce~s •. -from some samples would coincide or
fall closer to the population mean .
Quality _of Census Data ~hile those fiom · others would fall
far apart . ·
A dmsus comprises of a series
- of different - operations taking In ord.er to compare precision
place qne ~fter the other. ' Every of sample estimates of different
operation is likely _ to add dif- i terns, their sampling errors are
fe r ent type of errors to the end _ brought on comparable terms by
res ult. It is, therefore; not pos- -expressing them as percentages of
sible to have a linear· measure of relevant sample _estimates. The
the . total error introduced. resultant measure is designated as
However, ·the total errors likely to co~efficient of variation which cari
be - present in the _ sample census be estimated dir~ctly from the cen7
results - are probably classified sus data . For the 1990 Census data
into the following : co-efficient of variation have been
. '
worked out for a number of
a) • Sampling errors important items and given in
. ·~ . . Appendix- IV.
b) "·• Non-Sampling ' Errors
The formula used for estimat-
In- addition to the extent of ing rel-variances is described in
total errors likely to be present the preceding paragraphs . _. At the
in t he ~esults, an idea about the sub-division level , iel- variance
~ual ity of results could be formed can be estimated -simply by first
,by examinlng the census data : raising the sample -o'bservations of
each patwar circle fa . stratum level
i _) .Fqr - internal inconsistencies using the overal_l , raising factor
a.rid reasonableness. for each holdin~ but - excluding all
National Certainty holdings , and
·ay: ciqm~arison with other cor- then summing up tci obtain _a n esti-
respdriding seri~s. mate for .each half stratum as

(xxxiv)