Bad Data Means Bad Models! (Here’s Why)

[@chougaghil]

Fascinating hand gestures and voice rythm, like you were a baritone RichardOnData Excellent content, for beginners and reckless data scientists

[@RichardOnData]

Thank you, those two populations are what I'm going for!

[@AndrewMoMoney]

Clear, concise, and worth a watch. Thanks for breaking it down, Richard!

[@RichardOnData]

That's how I roll! Thanks Andrew.

[@aikimark1955]

Both objects, big and small, pull towards one another.

[@RichardOnData]

Oops, you are very correct. I'll add a card about that later. I think you get where I was going with that (we humans feel the pull more than the Earth and are drawn to it when we hop up in the air), but it was poorly and incompletely put on my part. Last time I took physics was when I was a freshman in college... clearly, it shows :P

[@aikimark1955]

@@RichardOnData Yes. I knew what you meant to say. I just felt like interjecting the comment from my Physics brain for the benefit of your viewers. Keep up the good work.

[@sonishakukreja6856]

So simple and powerful. Glad to follow you:)

[@RichardOnData]

Glad to have you!

[@linghong2609]

Hi Richard, thank you for the video! Could you explain more on "the lower the variance of the feature, the higher the variance of the estimated coefficient"? Does it also hold true for classification models?

[@RichardOnData]

In the regression context, the variance of the slope estimate increases as: 1) multicollinearity increases, 2) the variance of sigma-squared increases, 3) sample size decreases, or 4) the variance of the feature decreases. Think of it this way: it's easier to figure out the slope of a line from many, well-separated points, compared to from a few points smushed together. Alternatively think of it this way... as we observe more variation in the feature, we have more diverse data with which to explain the response. This will sometimes be the case with classification models, but can certainly vary by method.

[@LewiUberg]

Many of your videos are mostly you speaking. Have you considered uploding them as podcasts? I have been missing a good DS podcast:)

[@RichardOnData]

Yes, when Ken Jee and I did our collaboration (and he'll have the other part go up on his channel in the next coming weeks), we both talked about that idea, since both our parts went far longer than expected. I'm trying to roll more of these out in the months to come!

[@eugenemensah5738]

First to view!

[@RichardOnData]

Appreciate the dedication man!