WEBVTT

00:00:00.000 --> 00:00:08.707 position:50% align:middle
So let's take a look at an hypothetical scenario of
somebody taking 60 items and ended up failing the exam.

00:00:10.970 --> 00:00:20.388 position:50% align:middle
So the test starts, and based on this person's
answer, we continue to estimate his ability.

00:00:21.610 --> 00:00:25.832 position:50% align:middle
So the test keeps going and going.

00:00:26.320 --> 00:00:30.288 position:50% align:middle
And you can see that...I mean,
even though this is a hypothetical example,

00:00:30.288 --> 00:00:34.548 position:50% align:middle
but we try to use the data that mirror reality.

00:00:34.548 --> 00:00:40.931 position:50% align:middle
And you can see that the 95% confidence
interval gets narrower and narrower.

00:00:40.931 --> 00:00:44.459 position:50% align:middle
Our estimation is pretty precise.

00:00:44.459 --> 00:00:52.316 position:50% align:middle
So by the 60th item, we start evaluation
to see if this person is a passer or a failer.

00:00:52.316 --> 00:00:56.518 position:50% align:middle
Person is clearly a failer,
so the person didn't pass the exam.

00:00:56.518 --> 00:01:00.792 position:50% align:middle
And as you can see here, even though
this person really didn't do so good...

00:01:00.792 --> 00:01:06.694 position:50% align:middle
I mean, by the 15th item, you can see that, you know,
he never quite got up to the passing standard

00:01:06.694 --> 00:01:13.538 position:50% align:middle
after that, but because by the 15th item,
we haven't covered the test plan content enough,

00:01:13.538 --> 00:01:18.408 position:50% align:middle
we'll keep asking questions until the 60th,
and then we'll make a decision whether this person

00:01:18.408 --> 00:01:19.714 position:50% align:middle
passed or failed.