Original subtitles

Instructor: Before we go any further,

let's take a minute to discuss the previous situation.

If this was a real world situation,

you would've many many points,

potentially forming four clusters.

With the risk of oversimplifying the matter,

B could represent small, expensive apartments or ripoffs.

A would represent small, reasonably priced apartments.

D, big, reasonably priced apartments

and C would represent big, cheap apartments or bargains.

All else equal, what are we likely to observe usually?

Small apartments would be cheaper

and big apartments would be more expensive.

Maybe the rip-offs.

Were representing apartments in the city center

while the bargains apartments in the suburbs.

If we separate them from the rest,

we will be left with something that looks very familiar,

our good old regression.

And that's how different statistical methods

communicate with each other.

Now, what about the initial four cluster situation?

Clustering in this case could help us identify

omitted variable bias.

In this situation,

you could think about clustering as a method

for exploring the data and realizing that

one or more significant variables

have not been included in the analysis.

So instead of predicting price based solely on size,

we may need to include location

to get our better prediction.

Okay, hopefully this lecture was useful

not only for your clustering

but your data science understanding as a whole.

Thanks for watching.