1
00:00:00,005 --> 00:00:04,004
- [Instructor] Most of us have heard of double-blind studies

2
00:00:04,004 --> 00:00:07,009
but most data scientists don't work in a setting,

3
00:00:07,009 --> 00:00:12,003
or an industry where they get an opportunity to conduct them

4
00:00:12,003 --> 00:00:16,001
A more complete name is a randomized

5
00:00:16,001 --> 00:00:20,006
double-blind placebo control study.

6
00:00:20,006 --> 00:00:23,004
They are often described with phrases

7
00:00:23,004 --> 00:00:28,000
like "the best method," or "the gold standard,"

8
00:00:28,000 --> 00:00:31,004
or "least likely to be biased,"

9
00:00:31,004 --> 00:00:34,005
or "the most convincing."

10
00:00:34,005 --> 00:00:37,002
That can inspire feelings of frustration

11
00:00:37,002 --> 00:00:41,001
because many of us are competing with the reputation

12
00:00:41,001 --> 00:00:45,000
of a technique that is not available to us.

13
00:00:45,000 --> 00:00:47,007
We'll discuss alternatives in this course,

14
00:00:47,007 --> 00:00:51,000
so we will have options but let's first,

15
00:00:51,000 --> 00:00:54,000
get clear about what the features are,

16
00:00:54,000 --> 00:00:57,006
and why they are considered so attractive.

17
00:00:57,006 --> 00:01:00,000
What makes it a double-blind?

18
00:01:00,000 --> 00:01:03,007
In a double blind-study, not only the patient

19
00:01:03,007 --> 00:01:09,000
but also the clinician is unaware of the assignment.

20
00:01:09,000 --> 00:01:10,007
The assignment being whether or not

21
00:01:10,007 --> 00:01:15,001
the patient was given the treatment, or the placebo.

22
00:01:15,001 --> 00:01:18,006
Without blinding the patient's beliefs

23
00:01:18,006 --> 00:01:21,009
in the efficacy could impact the results.

24
00:01:21,009 --> 00:01:25,003
The purpose of double blinding is that the clinician

25
00:01:25,003 --> 00:01:30,002
could accidentally give subtle clues during the treatment.

26
00:01:30,002 --> 00:01:33,004
There's even something called triple blinding

27
00:01:33,004 --> 00:01:36,007
where even the data analysts, and researchers

28
00:01:36,007 --> 00:01:40,004
don't know which group is which.

29
00:01:40,004 --> 00:01:45,009
So let's recap with a list of these characteristics.

30
00:01:45,009 --> 00:01:50,006
There's a placebo group, there's randomization,

31
00:01:50,006 --> 00:01:55,004
it's prospective meaning that assignment to treatment,

32
00:01:55,004 --> 00:02:00,009
or a placebo is done before the study begins.

33
00:02:00,009 --> 00:02:05,006
And finally, as we've discussed, there is blinding.

34
00:02:05,006 --> 00:02:08,003
These characteristics work together

35
00:02:08,003 --> 00:02:11,008
to produce the type of study that most scientists,

36
00:02:11,008 --> 00:02:16,000
and statisticians find the most convincing.