The Tea Lady Experiment - Numberphile

Po-Ling Loh discusses hypothesis testing and a famous experiment involving cups of tea. More links & stuff in full description below ↓↓↓

NUMBERPHILE

Video by Brady Haran and Pete McPartlan

[joshcain]

Ah, so this is the student's tea-test I've been hearing so much about

[griof]

0:38 that is the smoothest hand-draw normal distribution I have ever witnessed

[zzzaphod]

Seems like the video ended right as it was about to get interesting - why not talk more about the graph problem?

[peterflom]

My favorite professor in grad school used to say "stop p-ing on the research".

[macronencer]

4:55 As a keen anagrammatist, I truly appreciated this!
Oh, by the way... "Resampling method" could be "The Random Glimpse" :)

[YouPlague]

If you separated the tea part itself as an explanation of P-value that would be a great tool for teaching. As it stands it's drowned by the intro.

[pascalbercker]

A missed opportunity to explain what is wrong with p-values and significance testing and the arbitrary nature of the 5% cut-off. This would dovetail into a brief excursion of the replication crisis. By the way, has this tea-tasting experiment ever been replicated?

[danpatterson]

If the tea lady knows how many of each type (milk first or tea first) are in the sample, that might affect her decision-making- say she has tasted six of the eight samples, and only labeled two as "tea first"- she would have a motive to label the last two "milk first" so as not to invalidate her own claim.

[c]

As my teacher used to say: "If the p is small, reject the Ho"

[pyroMaximilian]

This is the first numberphile video I got absolutely nothing from. There were a lot of gaps in the explanation and the end goal wasn’t clear.

[Kevlux]

Great explanation tying percentile to statistical significance and with a value we’re actually weighing against (that 5%). Very cool

[dibenp]

Aw snap. I was totally engrossed in learning about hypothesis testing as presented by Po-Ling Loh when the video ended. I'd love to watch an extended video. But thanks for the video we got. 👏

[DerekMitchell]

What about the situation with 8 cups, where you are not limited to 4/4. So all 8 could be milk first or 0 could be milk first. Where would the 5% cut off lie?

[olipito]

But then, did Muriel have superpowers or not?

[lucarnauddunoyer]

You gotta love when hypothesis and prediction are conflicted even at the highest level of sciences ... makes our lives harder when we try to teach students the right hypothesis (x influences y) vs prediction (the shape of that relationship) framework.

[Circuit_Woods_Tales]

I remember learning how to make those significance tests. Now I see why we actually need them—to prove wrong (or check) all those so-called psychics.

[eziola]

Love this! More videos with basic stats concepts explained please!

[TruthNerds]

off topic: Tannic acid and the high temperature of the tea can cause milk proteins to denature and coagulate (curdle). This could affect the taste. Whether the tea or the milk goes into the cup first can potentially influence this effect, but it depends on a variety of factors really.

For example, pouring the tea into the cup first and then adding milk might give the tea enough time to cool down a bit from contact with the (colder) cup, depending on how quick you are, so it might reduce the likelihood of curdling. It would seem to be more correlation than causation though regarding the order.

[mara_may]

9:41 it pains me that there are 9 permutations missing, lol. Its seems like there are 3 permutations on the bottom of each column that got cut off.

[neillane]

Really appreciate Dr. Loh's bringing prob/stat to the Numberphile. I'd add that for the height test, one might use logs to prevent the <0 problem.