WEBVTT

00:00:00.000 --> 00:00:01.520 align:middle line:90%


00:00:01.520 --> 00:00:05.280 align:middle line:84%
[INAUDIBLE] is a
quantitative method

00:00:05.280 --> 00:00:08.160 align:middle line:84%
and we must use
statistics, right?

00:00:08.160 --> 00:00:10.990 align:middle line:84%
Absolutely, and depending
on your research question,

00:00:10.990 --> 00:00:13.420 align:middle line:84%
there are a lot of different
choices to choose from.

00:00:13.420 --> 00:00:17.320 align:middle line:90%
OK, so which is the best?

00:00:17.320 --> 00:00:20.460 align:middle line:84%
I'm not sure it's correct
to frame it in terms of one

00:00:20.460 --> 00:00:22.352 align:middle line:90%
being better than the other.

00:00:22.352 --> 00:00:24.810 align:middle line:84%
It just depends on what's most
appropriate for the question

00:00:24.810 --> 00:00:25.530 align:middle line:90%
you're answering.

00:00:25.530 --> 00:00:27.900 align:middle line:84%
For example, if you're
interested in seeing

00:00:27.900 --> 00:00:30.990 align:middle line:84%
how the pupil size
differences evolve over time

00:00:30.990 --> 00:00:34.350 align:middle line:84%
then you're going to want
to use time series analyses.

00:00:34.350 --> 00:00:37.080 align:middle line:84%
However, a lot of the
time it's more than

00:00:37.080 --> 00:00:41.070 align:middle line:84%
sufficient to just average
over the entire trial window.

00:00:41.070 --> 00:00:43.260 align:middle line:84%
And, in fact, most of the
studies since the 1960s

00:00:43.260 --> 00:00:44.805 align:middle line:84%
have been doing
this sort of thing.

00:00:44.805 --> 00:00:47.610 align:middle line:84%
And in those cases, you want
to use things like t-tests

00:00:47.610 --> 00:00:49.755 align:middle line:90%
or repeated measures ANOVAs.

00:00:49.755 --> 00:00:51.520 align:middle line:90%
OK, standard statistics.

00:00:51.520 --> 00:00:52.020 align:middle line:90%
Sure.

00:00:52.020 --> 00:00:53.410 align:middle line:90%
How would they look like?

00:00:53.410 --> 00:00:55.480 align:middle line:90%
Well, I can show you.

00:00:55.480 --> 00:01:00.120 align:middle line:84%
So here we've got the raw data
that comes right out of the eye

00:01:00.120 --> 00:01:02.090 align:middle line:90%
tracker.

00:01:02.090 --> 00:01:05.239 align:middle line:84%
And you can see that we've got
a column for time, the subject

00:01:05.239 --> 00:01:06.170 align:middle line:90%
number, and the trial.

00:01:06.170 --> 00:01:08.640 align:middle line:84%
And then we have, obviously,
the pupil diameter,

00:01:08.640 --> 00:01:11.130 align:middle line:84%
which is going to be
very important here.

00:01:11.130 --> 00:01:12.950 align:middle line:84%
But, as is, this
isn't so useful.

00:01:12.950 --> 00:01:14.690 align:middle line:84%
We can't run the
statistics on this.

00:01:14.690 --> 00:01:19.190 align:middle line:84%
Instead we have to aggregate the
data and add our conditions in.

00:01:19.190 --> 00:01:24.730 align:middle line:84%
And now I have the average
pupil size here in this column.

00:01:24.730 --> 00:01:30.860 align:middle line:84%
And then we can run statistics
on that, which I've done here.

00:01:30.860 --> 00:01:34.930 align:middle line:84%
And as you can see, when we
run the ANOVA it gives us

00:01:34.930 --> 00:01:38.570 align:middle line:84%
two significant effects, one for
this condition called entropy.

00:01:38.570 --> 00:01:42.200 align:middle line:84%
And the other for the condition
aptly called condition.

00:01:42.200 --> 00:01:43.600 align:middle line:90%
OK, that sounds great.

00:01:43.600 --> 00:01:47.740 align:middle line:84%
But let's say I'm
studying music.

00:01:47.740 --> 00:01:50.770 align:middle line:84%
And I want to know what
happens to the pupil

00:01:50.770 --> 00:01:54.640 align:middle line:84%
at the particular moment
during a musical piece.

00:01:54.640 --> 00:01:58.520 align:middle line:84%
How should use
statistics in that case?

00:01:58.520 --> 00:02:00.190 align:middle line:84%
So this is one of
the examples of when

00:02:00.190 --> 00:02:02.480 align:middle line:84%
you're going to want to
use time series analysis.

00:02:02.480 --> 00:02:08.080 align:middle line:84%
So here I've plotted the pupil
size as it evolves over time.

00:02:08.080 --> 00:02:11.830 align:middle line:84%
On this x-axis here, you've got
time, pupil size on the y-axis

00:02:11.830 --> 00:02:12.520 align:middle line:90%
here.

00:02:12.520 --> 00:02:14.750 align:middle line:84%
Different lines represent
different conditions.

00:02:14.750 --> 00:02:19.780 align:middle line:84%
And as you can see here,
we've got within subject 95%

00:02:19.780 --> 00:02:21.130 align:middle line:90%
confidence intervals.

00:02:21.130 --> 00:02:23.590 align:middle line:84%
And it looks like there might
be some differences emerging

00:02:23.590 --> 00:02:25.150 align:middle line:90%
within this time window.

00:02:25.150 --> 00:02:26.980 align:middle line:84%
And so then, maybe
we want to look

00:02:26.980 --> 00:02:31.770 align:middle line:84%
at that area for future
analysis and statistics.

00:02:31.770 --> 00:02:35.730 align:middle line:84%
Great, so we need to be a
bit flexible about which

00:02:35.730 --> 00:02:41.040 align:middle line:84%
particular kind of statistics
applies to a specific study.

00:02:41.040 --> 00:02:44.340 align:middle line:84%
Absolutely, statistics
are just like a tool kit.

00:02:44.340 --> 00:02:46.920 align:middle line:84%
You're not going to use a
hammer to screw in a screw,

00:02:46.920 --> 00:02:48.400 align:middle line:90%
for example.

00:02:48.400 --> 00:02:51.420 align:middle line:84%
You could, but it wouldn't
be the best way to do it

00:02:51.420 --> 00:02:52.970 align:middle line:90%
Right.

00:02:52.970 --> 00:03:03.000 align:middle line:90%