Exploring the effects of changing sample sizes

 

Syntax

BD>varysize : E1,E2,E3

where E1, E2, and E3 are any valid equations which result, on rounding, in positive integers , such that and .

The VARYSIZE:  command is used for the following purpose. Whenever an exchangeable adjustment has taken place (that is, an adjustment satisfying the conditions of §16.2) then it is possible to explore the effect, on the current adjustment, of changing the sample sizes in the sense that some of the output generally available for an adjustment is generated for changed sample sizes by exploiting the underlying exchangeability. That is, changing the sample size generally leads to an adjustment quite different to the current one, but certain summary aspects of the altered adjustment may be obtained cheaply.

Assume that the initial adjustment was for [B/C(n)]. The effect of this command is as follows. Any output currently selected by the options summarised below is computed and given assuming the sequence of adjustments , , and so forth; and finally for the largest integer m such that , . The options which are affected are as follows:

bases 

for titling, and to summarise the elements in the adjustment;

ru 

for a summary of the effect of the adjustment on the system resolution;

v 

for output of the adjusted variances;

e 

for output of the adjusted expectations;

e* 

for output of the standardised adjusted expectations;

pv 

for output of partial adjustment variances, if any.

Note that any comparative information yielded refers to comparisons between the current and previous adjustment. In the context of the VARYSIZE:  command, results are output for different sample-sizes on effectively the same adjustment, and the previous adjustment remains unchanged.