14.03.2015 Views

Behavioural Surveillance Surveys - The Wisdom of Whores

Behavioural Surveillance Surveys - The Wisdom of Whores

Behavioural Surveillance Surveys - The Wisdom of Whores

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

One thing to keep in mind with truck<br />

drivers, is that they are moving all the time.<br />

So depending on the distances traveled,<br />

sampling from several different locations<br />

might tend to result in a lot <strong>of</strong> duplication.<br />

<strong>The</strong>refore, if there is a single location (such as<br />

a bridge or weighing station), that all drivers<br />

must pass through to get where they are<br />

going, it is sometimes a better idea for the<br />

interviewing team to remain stationary.<br />

<strong>The</strong> strategy would be to set up multiple<br />

time-location units at the same spot, but on<br />

different days <strong>of</strong> the week, and times <strong>of</strong> the<br />

day, over a period <strong>of</strong> several weeks.<br />

By repeatedly sampling from a few sites,<br />

it might be possible to more efficiently<br />

“capture” the population as they pass through.<br />

Guidance on selected sampling<br />

problems<br />

Despite careful up-front research and<br />

planning, surprises <strong>of</strong>ten occur during the<br />

course <strong>of</strong> survey fieldwork. Guidance on<br />

what should be done when several samplingrelated<br />

surprises arise is provided below.<br />

When insufficient numbers <strong>of</strong> the survey<br />

population members are found at sample sites<br />

In many <strong>of</strong> the sampling schemes described<br />

above, the number <strong>of</strong> sites to be chosen will<br />

be based upon an “expected” number <strong>of</strong><br />

sub-population members to be found per site<br />

during a specified time interval. What should<br />

be done if it is found during the course<br />

<strong>of</strong> fieldwork that the actual number <strong>of</strong><br />

sub-population members found is substantially<br />

lower than expected ?<br />

<strong>The</strong> preferred course <strong>of</strong> action is to have<br />

anticipated such a possibility when determining<br />

survey sample size requirements, and to<br />

have adjusted the target sample size upward<br />

as a precaution against “surprises” in the field.<br />

Such allowances for non-response are typically<br />

made in sample size calculations for household<br />

surveys. Given that the number <strong>of</strong> sub-populations<br />

members expected to be found per site<br />

will <strong>of</strong>ten be no more than an informed guess,<br />

a strong case may be made for following the<br />

same practice in all sub-population surveys.<br />

<strong>The</strong> recommended procedure is to increase<br />

the number <strong>of</strong> sites to be included in the<br />

sample by 10% or so. So, for example, if it<br />

were determined that 30 PSU’s would be<br />

covered in a given survey, the recommended<br />

course <strong>of</strong> action would be to increase this<br />

to 33, 35, or even 40 if resources permit.<br />

Although returning to sample sites for an<br />

additional period <strong>of</strong> data collection is an<br />

option, this option is less desirable than<br />

increasing the number <strong>of</strong> sample PSU’s chosen<br />

for two reasons. First, if the reason that<br />

insufficient numbers <strong>of</strong> sub-population members<br />

were found is that the estimates <strong>of</strong> the expected<br />

daily volume <strong>of</strong> sub-population members were<br />

faulty (i.e., too low), returning to the same<br />

sites will be a relatively inefficient way <strong>of</strong><br />

increasing the sample size. Secondly, sampling<br />

additional PSU’s will serve to increase the<br />

precision <strong>of</strong> survey estimates more so than<br />

sampling additional cases per PSU. Returning<br />

to sample PSU’s different numbers <strong>of</strong> times<br />

also complicates the calculation <strong>of</strong> sampling<br />

probabilities. If this option is chosen, the<br />

number <strong>of</strong> return visits made to each site<br />

needs to be carefully documented.<br />

B EHAV I OR A L S U R V EI L L A NC E SURV EY S APPEN DI X 3<br />

335

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!