All language subtitles for 03 - Consider the limits of hypothesis testing

- [Instructor] As much as we like to think

that statistics is an exact science,

in fact, there's a lot of room for uncertainty.

In this movie,

I want to go over some of the limits of hypothesis testing

to give you some cautions about your data and your analysis.

The first is that,

even if your analysis meets the 95% standard

that most analysts use,

you will still be wrong one out of 20 times.

And if you've ever performed any type of analysis

or even flipped a coin

and gotten four or five tails in a row,

then you know how likely it is that events can happen

one out of 16, one out of 20, or even one out of 100 times.

Also, you might run into errors

in data collection or analysis

that can skew your results.

Some errors in data collection

are not randomized during your samples

or using old data.

So make sure that your data is collected properly

and that it isn't too old

and that circumstances haven't changed

to make the analysis invalid.

And also, remember that correlation is not causation.

Just because two things happen together

doesn't mean that one caused the other.

That said, correlation does tell you where to start looking.

If two things happen together

and it's plausible that there is a way

that one could have caused the other,

then that's a great place to start looking.

As always, this is where your intuition

and knowledge of the business will help out a lot.