Tip:
Highlight text to annotate it
X
(male narrator) So next, we're gonna look at some, uh...sources of bias.
Uh...bias is a-a systematic distortion
of your statistical results,
which come from something influencing the data.
Now one version of this, we sort of already talked about,
and that would be what's called a "sampling bias."
Uh...and sampling bias is what you get
when your sample is not actually representative
of your population, uh...and then the data
you collect from your sample can't really be trusted.
And it has a... it's distorted,
because it is not representing the population
that you're actually trying to talk about.
Let's look at some other cases.
So consider a recent study, which found that chewing gum
may help raise math grades in teenagers.
This study was conducted
by the Wrigley Science Institute--
a branch of the Wrigley Chewing Gum Company.
Uh...what potential source of bias
should we be concerned about here?
And you're probably jumping on the fact right away
that this is saying gum is good,
uh...from a study done by a gum company.
Uh...this is called... a...uh...
or our concern here is a self-interest study.
This is called a self-interest study,
where the person conducting the research
has an interest in the result--
uh...in this case a financial interest.
Now that does not necessarily mean
that the data is invalid, or that the study is invalid.
There has been a lot of very valid research done
by companies with an interest in the study.
Certainly, all pharmaceutical research is done
by, uh.. a self-interested party.
But it certainly means that we should look at the results
with an additional level of skepticism
and really dig down to make sure that the results
that they're providing are valid.
So suppose we have a survey that asks people,
when was the last time you visited the doctor?
Uh...what should we be concerned about here?
And the big concern here is something called...
something called "response bias."
Something called response bias, and this is when the responder,
uh...gives inaccurate responses for any of a variety of reasons.
In this case, it's probably, uh...a memory issue,
where, you know, somebody might not remember,
uh...the last time they visited the doctor
and might think it was, you know, two months ago,
when it was, in fact, you know, seven months ago.
And so this is a bias.
Now some of these are sort of intentional nefariousness.
Uh...you know, like a self-interest study,
we might be concerned
about intentional, uh...you know, deception.
But oftentimes, these biases are not intentional,
um...and a response bias is certainly one of those.
So suppose that we...a survey asks participants a question
about their interactions with members of other races.
You know, how they look at... how well they get along
with, uh...you know, black people or Asians.
Um...what would we be concerned about here?
Now here we have something called
a perceived...perceived lack of...anonymity.
In other words...
if, uh...particularly if this was being given
in a face-to-face, um...uh...interview process,
the person who's being asked might,
um...particularly if they have, you know, racial biases,
they might, uh... be uncomfortable sharing that,
because they don't wanna be perceived as racist.
Uh...and so they may, um...be uninclined
to given, uh...to give an accurate answer.
So now suppose a survey asks,
uh...do you support funding research
of alternative energy sources
to reduce our reliance on high-polluting fossil fuels?
Uh...so the sort of the bias issue here
is something called a "loaded or leading question."
Uh...so this question is considered
to be a loaded question, because the wording--
this "high-polluting fossil fuels"--
uh...sort of leads the respondent
towards an answer.
I mean, you read this, and you say,
"Oh, high-polluting fossil fuels.
I don't like those."
Uh...so, "Yeah, of course I support funding research."
Uh...whereas, if the question said something like,
"Do you support funding research of alternative energy sources,
"which may include... uh...which will be funded
through increases in gas taxes?"
Uh...chances are a lot more people
would say, no, to that question.
So let's look at the next one here.
A telephone poll, uh... to ask...uh...asks the question,
uh...do you...do you often have time
to relax and read a book?
And 50% of the people called refused to answer the survey.
Uh...so the issue here is something called...
uh...oops...it's called "non...response...bias."
In this case, the issue is that people
aren't responding to the question,
and when we...particularly when the question is like,
"Do you have time to relax?"
If half the people are not even answering the question,
uh...chances are this is going
to influence the accuracy of your study.
'Cause the people who answer are probably gonna be the people
who actually have time to answer studies and surveys like this.
So one last one.
To determine how long it takes to hit the brakes
when an animal runs in front of the car,
uh...100 college students are recruited
and put through a simulator.
So the issue here goes back to where we started.
This is what's called "sampling bias."
And it really depends upon who our target population was here.
But if we're really figuring out...
trying to figure out how long it takes people--
where this is just sort of a generic people--
uh...then using college students as our sample,
uh...is probably gonna slant the results.
Because college students are probably gonna have
a higher, quicker reaction time
than, uh...than some other Americans,
such as, um...you know, senior citizens.
And so...this probably... this sample
is probably not representative of the population
and is probably going to skew those results.