3 Conservative bias correction for the Horvitz-Thompson variance estimator under non-measurability
Peter M. Aronow and Cyrus Samii
Previous | Next
The case where may be less than zero for a non-measurable
design suggests the need for some adjustment that will guarantee a bias that is
weakly bounded below by zero. We first develop a general bias correction that
is guaranteed to be conservative, later providing a special simplified case for
practical usage.
3.1 General
formulation
Consider the following variance estimator:
where are positive real numbers such that for all pairs with The estimator is guaranteed to produce an
expected value greater than or equal to the true variance for all designs, and
is thus conservative. We state this property formally:
Proposition 2. The expected value of
Proof. By Young's inequality,
if Define such that,
and
Therefore
The associated Horvitz-Thompson estimator of would be
which is unbiased by and
Since by Proposition 1,
Substituting terms,
is justified as a conservative estimator for
the case when is not known to be positive. This estimator is
unbiased under a special condition:
Corollary 1. If, for all pairs such that (i) and (ii)
Proof. By (i), (ii) and Young's
inequality,
Therefore,
It follows that and
If any units are in clusters (i.e., these units should be totaled into one larger
unit before estimation. Combining units will tend to reduce the bias of the
variance estimator because only pairs of cluster-level totals will be included
in as opposed to all constituent pairs.
3.2 Simplified
special case
In general, it would be difficult to assign optimal
values of and for all pairs such that Instead, we examine one intuitive case,
assigning all
As a special case of is also conservative:
Corollary 2. The expected value of
Proof. For all pairs such that Proposition 1 therefore holds.
The choice to set all is justified by the fact that it will yield
the lowest value of the estimator subject to the constraint that and are fixed as constants and over all
Corollary 3. Among the class of estimators defined as the set of estimators such that all and all
Proof. By simple algebra,
By Young's inequality, given but equality must hold if Similarly, but equality must hold if Since and any choice can only yield
Given all values of and it is possible to derive an optimal vector of and values that varies over but such a derivation may not be of practical
value.
Previous | Next