WEBVTT

00:00:00.000 --> 00:00:05.380 align:middle line:90%


00:00:05.380 --> 00:00:07.080 align:middle line:84%
So I'm collecting
a variety of data

00:00:07.080 --> 00:00:11.040 align:middle line:84%
on lots of different animals at
the moment in the laboratory,

00:00:11.040 --> 00:00:15.490 align:middle line:84%
including small protists
in an aquatic environment,

00:00:15.490 --> 00:00:19.260 align:middle line:84%
looking at how their
movement rates change,

00:00:19.260 --> 00:00:20.880 align:middle line:84%
and how their
metabolic rates change

00:00:20.880 --> 00:00:22.410 align:middle line:90%
with different temperatures.

00:00:22.410 --> 00:00:26.520 align:middle line:84%
We're looking at wood lice
and the effects of huddling

00:00:26.520 --> 00:00:28.380 align:middle line:84%
in woodlice on their
metabolic rates,

00:00:28.380 --> 00:00:30.450 align:middle line:84%
and how humidity
affects that.

00:00:30.450 --> 00:00:34.020 align:middle line:84%
I also collect data
directly from the internet.

00:00:34.020 --> 00:00:37.650 align:middle line:84%
So collating data from
other people's papers,

00:00:37.650 --> 00:00:41.850 align:middle line:84%
or even collecting data through
YouTube videos, on the movements

00:00:41.850 --> 00:00:44.205 align:middle line:84%
and speed of movements
of different animals.

00:00:44.205 --> 00:00:49.470 align:middle line:90%


00:00:49.470 --> 00:00:52.410 align:middle line:84%
If you try to answer a
large number of questions

00:00:52.410 --> 00:00:55.080 align:middle line:84%
with the same data
set, particularly

00:00:55.080 --> 00:00:59.497 align:middle line:84%
questions you weren't initially
using that data set to answer,

00:00:59.497 --> 00:01:01.080 align:middle line:84%
it is more and more
likely that you're

00:01:01.080 --> 00:01:05.430 align:middle line:84%
going to uncover a
significant result.

00:01:05.430 --> 00:01:07.740 align:middle line:84%
And more likely the
result is actually

00:01:07.740 --> 00:01:10.920 align:middle line:84%
not truly a significant
result. So it's

00:01:10.920 --> 00:01:12.630 align:middle line:84%
important to set
out what one wants

00:01:12.630 --> 00:01:14.520 align:middle line:90%
to do the study, first of all.

00:01:14.520 --> 00:01:18.090 align:middle line:84%
Then there's the added
issue of removing some data.

00:01:18.090 --> 00:01:22.403 align:middle line:84%
If there's an argument that it's
a bad data point, or data set.

00:01:22.403 --> 00:01:24.570 align:middle line:84%
Now obviously, that can be
true for certain reasons.

00:01:24.570 --> 00:01:27.690 align:middle line:84%
But we have to be very,
very careful about why

00:01:27.690 --> 00:01:28.890 align:middle line:90%
we want to remove data.

00:01:28.890 --> 00:01:34.290 align:middle line:84%
And as much as possible, have an
objective reason for doing so.

00:01:34.290 --> 00:01:37.140 align:middle line:84%
In recognition that as human
beings, we are fallible

00:01:37.140 --> 00:01:39.090 align:middle line:84%
and that you know,
it's quite easy for us

00:01:39.090 --> 00:01:42.060 align:middle line:84%
to want to move a data set
in a certain direction.

00:01:42.060 --> 00:01:46.470 align:middle line:90%


00:01:46.470 --> 00:01:47.970 align:middle line:84%
I definitely consider
different ways

00:01:47.970 --> 00:01:50.790 align:middle line:84%
of interpreting my findings
because every data set is

00:01:50.790 --> 00:01:52.440 align:middle line:90%
so different from one another.

00:01:52.440 --> 00:01:55.290 align:middle line:84%
And I find that even
throughout the start of my PhD,

00:01:55.290 --> 00:01:57.330 align:middle line:84%
towards the end
of it, that what I

00:01:57.330 --> 00:01:59.550 align:middle line:84%
was exposed to previously
and the type of data

00:01:59.550 --> 00:02:01.772 align:middle line:84%
I was working with
changed over time.

00:02:01.772 --> 00:02:04.230 align:middle line:84%
And that meant that I had to
keep going back to the drawing

00:02:04.230 --> 00:02:06.150 align:middle line:84%
board to find out
what are the newest

00:02:06.150 --> 00:02:10.530 align:middle line:84%
and maybe more sophisticated
data analyses happening

00:02:10.530 --> 00:02:12.060 align:middle line:84%
and how can that
apply to my data?

00:02:12.060 --> 00:02:15.870 align:middle line:84%
How can I best
describe what I found?

00:02:15.870 --> 00:02:18.320 align:middle line:84%
And that changes as
everything else does.

00:02:18.320 --> 00:02:20.820 align:middle line:84%
So I think it's very
important to keep up to date

00:02:20.820 --> 00:02:23.430 align:middle line:84%
and make sure that
however you interpret it

00:02:23.430 --> 00:02:26.222 align:middle line:84%
ot however you analyse it,
it changes, according

00:02:26.222 --> 00:02:27.430 align:middle line:90%
to the sciences around you.

00:02:27.430 --> 00:02:32.760 align:middle line:90%


00:02:32.760 --> 00:02:34.700 align:middle line:84%
Well, bias is one of the
most important things

00:02:34.700 --> 00:02:36.510 align:middle line:84%
that we have to
consider in research.

00:02:36.510 --> 00:02:38.810 align:middle line:90%
Bias is unavoidable.

00:02:38.810 --> 00:02:41.570 align:middle line:84%
In any research, our
methods of analysis

00:02:41.570 --> 00:02:43.470 align:middle line:90%
are not going to be perfect.

00:02:43.470 --> 00:02:45.560 align:middle line:84%
So we have to
consider bias, ideally

00:02:45.560 --> 00:02:48.380 align:middle line:84%
since the planning
of the research,

00:02:48.380 --> 00:02:51.710 align:middle line:84%
since we start with the first
ideas about what we want to do.

00:02:51.710 --> 00:02:55.820 align:middle line:84%
If we consider bias then, then
we can improve our methods.

00:02:55.820 --> 00:02:56.810 align:middle line:90%
We can correct.

00:02:56.810 --> 00:03:00.560 align:middle line:84%
We can do the
necessary adjustments.

00:03:00.560 --> 00:03:03.950 align:middle line:84%
So in our case, there
was a bias for us,

00:03:03.950 --> 00:03:06.500 align:middle line:84%
for example, with the
traps that we use.

00:03:06.500 --> 00:03:09.510 align:middle line:84%
These traps our for specific
species of mosquitoes.

00:03:09.510 --> 00:03:11.030 align:middle line:84%
So we knew that
there was not going

00:03:11.030 --> 00:03:16.250 align:middle line:84%
to be a complete representation
of the community of mosquitoes,

00:03:16.250 --> 00:03:17.490 align:middle line:90%
in that sense.

00:03:17.490 --> 00:03:21.540 align:middle line:84%
So we had to complement our
trapping with other methods.

00:03:21.540 --> 00:03:23.990 align:middle line:84%
So we also sample
immature mosquitoes

00:03:23.990 --> 00:03:25.520 align:middle line:90%
in the breeding sites.

00:03:25.520 --> 00:03:27.920 align:middle line:84%
So that was also
important for compensating

00:03:27.920 --> 00:03:31.300 align:middle line:90%
for this bias in the traps.