At risk of being (pretty rightly)
accused of getting involved in something about which, in
detail, I know very little (I'm a very amateur bird-watcher,
though I do have qualifications in science and economics of
some relevance to the question I'm asking!), I would be
interested to know something about the statistical
"reliability" of the data shown in the last few days on rufus
and golden whistler population time series trends and
population seasonalities. My sense is that only if the data
are statistically sound can reasonably reliable general
conclusions be inferred from them. (I'm not for one moment
suggesting there is a "reliability
problem" - it's just that I've not seen any comment on the
question, either way).
I am wondering (a) what sample
sizes are these data based on - essentially how many actual
bird sitings and how many "nil" sitings from how many
observation sites/areas are represented by each data point on
the graphs; and (b) what proportion of the "area/region of
interest" (eg, the ACT if it's trends in the ACT that are
being reported) - or perhaps more usefully what proportion of
the species' typical habitat' in (say) the ACT, was included
in the aggregate data?
I ask these questions, in part
due to prior experience in ostensibly scientific data and
analyses on Southern Bluefin Tuna (SBT) population, when I was
involved as a policy adviser in trying to interpret the
analyses from what had turned into a major dispute between
Australian and Japanese fish population scientists over SBT
populations and "responsible" fishing quotas.
To simplify the detail of the
analysis, the dispute turned on significant differences in the
two countries' scientists' methodologies in estimating (and
projecting) current and future SBT populations. In part, an
element in the different analyses and conclusions was the
extent to which population data from a set of "sample areas"
which necessarily covered only a small proportion of the total
range of SBT could, or could not, be regarded as
"representative".
So, to put my questions that way,
can you be confident that the whistler (both species) sample
areas used in the data/analysis that have are now being
presented and discussed are "representative" enough , and the
number of sitings made (whether positive or not) large enough
to enable reliable inferences to be drawn?
Thank you.
