All language subtitles for 01 - Formulate a hypothesis

- [Instructor] When you analyze your data using statistics,

you should know what aspect of your data you want to test.

The statement that embodies your test

is called a hypothesis.

Your hypothesis is an educated guess

about the characteristics of a data set

and the circumstance it describes.

For example, you could create a hypothesis that says

the state a customer lives in

is related to the amount of olive oil

they order from your company.

There are two parts of a hypothesis

used for hypothesis testing.

The first is the null hypothesis,

which simply says that factor A has no effect on B.

In the example I gave earlier,

it would say that the state a customer lives in

has no effect on how much they spend at my store.

The alternative hypothesis says factor A affects B.

So the state that a customer lives in

does affect how much they spend at my store.

An alternative hypothesis

can be directional or non-directional.

A directional hypothesis looks at either a value

that is greater than or less than a target.

For example, customers who were part

of a specific marketing group

might spend more than the average

based on the effect of that advertising,

whereas customers who are previous customers

but had not ordered for a while would spend less.

A non-directional alternative hypothesis

looks for a comparison to a specific value.

You might say that 30% of customers who ordered one product

would order another.

And you're looking for a variation either above or below.

So it can be both instead of just one.

So, how do you create an effective alternative hypothesis?

Well, first, you need to state

that there is a relationship between the variables

and then base your alternative

on your knowledge of the world.

For example, if you believe that your advertising campaign

will make a difference in sales,

you need to state that.

You should express your hypothesis simply and briefly

and also make sure you can test your alternative hypothesis.

Your hypothesis needs to be based on data

that you are able to collect

and that you believe should be reliable,

so that you can perform an accurate test.