9. SAMPLING
Details of the size and nature of the sample.
How offices and cases were selected
9.1 The Benefit Review pilot of FAMC was designed
to test the review methodology and was not intended to produce
conclusive statistically valid results. For this reason ASD advised
that a sample of 300 would be sufficient to test the methodology
and inform a decision on the sample size that would be required
for a main study.
Nature of the Sample
9.2 In order to obtain reliable indications of
any difficulties in dealing with claims from the self-employed,
which make up only about 14 per cent of FAMC claims, it was necessary
to over-sample self-employed cases. The sample was therefore stratified
to include 200 employed and 100 self-employed cases.
9.3 In order to facilitate the detection of incorrectness
it was decided to visit cases as close to the date of adjudication
as possible. The standard approach to Benefit Review sampling
is to select the sample in one batch, the FIs then having between
6 and 11 weeks to complete their visits. The drawback of using
this approach with FAMC is that by the time the FI was nearing
the end of the fieldwork his cases would no longer be "fresh"
and it was felt this would reduce the chances of finding incorrectness.
The sample was therefore selected in three batches, in weeks 0,
4 and 8 of the fieldwork, to ensure that FIs had a steady supply
of "fresh"
cases to visit.
Selection of Offices
9.4 The liveload of FAMC is approximately 725,000.
A selection of 300 cases across the whole country would have produced
too wide a geographical spread of customers to make visiting cost-effective
so it was decided to draw the sample from 15 locations, allowing
20 cases from each. Statisticians randomly selected 15 offices
and FAMS were asked to nominate an investigator from each office.
Selection of Cases
9.5 All paperwork relating to FAMC claims, once
actioned, is kept at Filestore at Nelson. No clerical records
are stored at the FAMC Unit.
9.6 ASD arranged for FAMC Computer System (FCCS)
Support to run three separate scans of the FAMC database, one
for each of the three sample batches, at 4 week intervals. ASD
provided FCCS with postal districts and postal areas for the 15
sample offices and all claims bearing those post-codes and having
a common adjudication date, within a one week window, were listed.
Also listed were cases without postcodes.
9.7 The lists were passed to ASD who allocated
postcodes to non-postcode cases. This was done in two stages.
In the first stage the cases were run through a software package,
Quick Address. This allocated the majority of postcodes. In stage
2 the remainder were allocated manually using a gazetteer. Once
all cases had postcodes allocated the requisite number of cases
was randomly selected for each of the 15 sample offices.
|