Appendix
Jae Kwang Kim and Shu Yang
Previous | Next
A.1 Replication
variance estimation
For variance estimation, replication methods
can be used. Let
be the
replication weights such that
is consistent for the variance of
where
is the replication size,
is the
replication factor that depends
on the replication method and the sampling mechanism, and
is the
replicate of
In delete-1 jackknife variance
estimation,
and
To apply the replication method in FFI, we
first apply the replication weights
in (2.4) to compute
Once
is obtained, we use the same
imputed values to compute the initial replication fractional weights
with
The variance of
computed from (3.4), is then
computed by
where
comes from solving
and
is defined in (A.1).
We now discuss replication variance estimation
of the FHDI estimator
computed from (3.8). Define
if
and
otherwise. Note that
is computed via two steps: in the
first step, a systematic PPS sampling is used with the selection probability
proportional to the fractional weights from the FFI method. In the second step,
the calibration weighting method using the constraint (3.5) with
is used. Thus, the replicate
fractional weights are also computed in two steps. Firstly, the initial
replication fractional weight for
is then given by
where
is the fractional weight for FFI
defined in (2.6) and
is the
replication fractional weight
for FFI defined in (A.1). Secondly, the replication fractional weights are
adjusted to satisfy the calibration constraints. The calibration equation for
replication fractional weights corresponding to (3.5) is then
and
Either regression weighting or
entropy weighting can be used to obtain the replication fractional weights
satisfying the constraints. Once the replicate fractional weights are obtained,
the replicate estimate
is computed by solving
The replication variance estimator of
computed from (3.8), is given by
Because
is a smooth function of
the consistency of
follows directly from the
standard argument of the replication variance estimation (Shao and Tu 1995).
A.2 Proof of Equation
(4.5)
Using
where
Based on Taylor linearization and
the fact of (4.4), we have
If we know the true density, the correct
fractional weights in (3.3) can be expressed by
where
and
So,
and
which proves (4.5).
A.3 Extension to a
non-ignorable missing case
We consider an extension of the proposed method
to a non-ignorable missing case. Under the non-ignorable missing assumption,
both the conditional model
and the response probability
model
are needed to evaluate the expected
estimating function in (4.6). Let the response probability model be given by
, for some
with a known
function. We assume that the
parameters are identifiable as discussed in Wang, Shao and Kim (2013).
In PFI, according to Kim and Kim (2012), the
MLE
can be obtained by solving
and
where
and the fractional weights are
given by
The solution to (A.4) and (A.5) can be obtained
via the EM algorithm. In the EM algorithm, the E-step computes the fractional weights
in (A.6) using the current parameter values and the M-step updates the
parameter value
and
by solving
and
In the proposed FFI method, the fractional
weights are given by
with
Because
The fractional weights can be computed from
with
Thus, we can use the following EM algorithm to
obtain the desired parameter estimates.
(I-step) For
each missing unit
take
imputed values as
from
where
(E-step) The
fractional weights are given by
and
(M-step) Update
the parameter
and
by solving the following imputed
score equations,
and
Note that the I-step does not have to be
repeated in the EM algorithm. Once the final parameter estimates are computed, the
fractional weights are computed by (A.8), which serve as the selection
probabilities for FHDI with a small imputation size
The same systematic PPS sampling
method as discussed in Section 3 can be used to obtain FHDI.
Previous | Next