Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Marshmallow Test: What Does It Really Measure?

theatlantic.com

Post reply on HN
Print view View original

271–280 of 282 posts

Re: The Marshmallow Test: What Does It Really Measure?

#271

Post by Retric »

ItsMe000001 wrote: post #266
Retric wrote: post #256

Earlier quoted context omitted.

I think you are failing to grasp the price differences and my argument. You can become obese eating rice and beans for less than the 1.50$ per day. Morbid obesity on cheap calorie dense food is cheaper than a healthy diet. This trend continues if you look at fast food or any quick options. So, no being fat is not cost prohibitive, but eating veggies is . Further, as I said if you eat empty calories like Purple Drank…

> I think you are failing to grasp the price differences and my argument. Thank you very much! I think I DO understand your argument, but I have the suspicion that you fail at understanding my very first reply already, which sums it up quite nicely. Your "argument" is none because it addresses an issue that does not exist. You countered something that you yourself invented, an issue that does not exist. Please, just…

Yea, I read what you said as they don't need to eat that many calories, but over time their body says they do.

Do you know any poor people? People that grew up in poor families? Just like everyone else when overweight their body wants to maintain that weight. That's going to take food, and in some cases a lot of it which is a real cost.

Eating cheap calorie dense food can cover based nutrition requirements. It's not optimal for health, but it's cheap, really really cheap.

Poor people also indulge in nutrition free food like soda when they can afford it, but then they still need to cover nutrition and the cheapest way to do that is high caloric foods. After a while they need to eat just as much to maintain that weight. But, at no point in that viscous cycle are veggies more than a garnish.

Would a lifetime of 3 servings of veggies a day help avoid weight gain while allowing for some indulgences? Of course, but it would also cost significantly more especially when raising a family which starts up the cycle again.

Re: The Marshmallow Test: What Does It Really Measure?

#272

Post by your-nanny »

SketchySeaBeast wrote: post #262
Retric wrote: post #256

Earlier quoted context omitted.

I think you are failing to grasp the price differences and my argument. You can become obese eating rice and beans for less than the 1.50$ per day. Morbid obesity on cheap calorie dense food is cheaper than a healthy diet. This trend continues if you look at fast food or any quick options. So, no being fat is not cost prohibitive, but eating veggies is . Further, as I said if you eat empty calories like Purple Drank…

Veggies are also perishable. - Rice and Beans don't go bad. You need to have a plan on what to do with said veggies when you get them and need to make sure you use them before they go bad. Rice and beans are as simple as two pots (one if you're lazy), a little bit of water, and 20 minutes of barely looking at the food.

A lot of beans take longer than 20 mins.

And beans are super nutritious.

Re: The Marshmallow Test: What Does It Really Measure?

#273

Post by eanzenberg »

xapata wrote: post #270
ethbro wrote: post #185

Earlier quoted context omitted.

* Above is true, if and only if: - Willpower is non-socially inherited - Willpower is proportionally developed (to their final amount) in children at the time of testing I would assume the hypothesis of the researchers was that children were more tabula rasa at the age tested, and therefore controlling for variables wouldn't be circular.

That's not the issue. Controlling for a variable is a model like `y ~ x + z` where x is the explanatory variable and z is the control. The commenter complained that since `y ~ z` and `x ~ z` the interpretation of the effect is strange. Not so. This is the sole purpose of controlling for confounders. If parents income were either not correlated with willpower or child's income, it could be dropped from the model. The…

If x and z are strongly correlated you can't determine causality for either within linear models. They are practically interchangeable and removing one will not decrease the effect of the other. "Controlling for", or keeping both on the right side, will decrease the effect of one over the other in unexpected ways, and why strong collinearity yields strange results in linear models.

Re: The Marshmallow Test: What Does It Really Measure?

#274

Post by xapata »

eanzenberg wrote: post #273
xapata wrote: post #270

Earlier quoted context omitted.

That's not the issue. Controlling for a variable is a model like `y ~ x + z` where x is the explanatory variable and z is the control. The commenter complained that since `y ~ z` and `x ~ z` the interpretation of the effect is strange. Not so. This is the sole purpose of controlling for confounders. If parents income were either not correlated with willpower or child's income, it could be dropped from the model. The…

If x and z are strongly correlated you can't determine causality for either within linear models. They are practically interchangeable and removing one will not decrease the effect of the other. "Controlling for", or keeping both on the right side, will decrease the effect of one over the other in unexpected ways, and why strong collinearity yields strange results in linear models.

If multicollinearity were such an issue, you'd be rejecting a wide range of techniques, such as including a lag term as a regressor for autoregressive models.

In fact, you could consider this situation as an autoregressive process, using generations instead of traditional time steps.

Re: The Marshmallow Test: What Does It Really Measure?

#275

Post by mlthoughts2018 »

xapata wrote: post #274
eanzenberg wrote: post #273

Earlier quoted context omitted.

If x and z are strongly correlated you can't determine causality for either within linear models. They are practically interchangeable and removing one will not decrease the effect of the other. "Controlling for", or keeping both on the right side, will decrease the effect of one over the other in unexpected ways, and why strong collinearity yields strange results in linear models.

If multicollinearity were such an issue, you'd be rejecting a wide range of techniques, such as including a lag term as a regressor for autoregressive models. In fact, you could consider this situation as an autoregressive process, using generations instead of traditional time steps.

Multicollinearity is a huge issue when fitting autoregressive and distributed lag models. In fact, it is even mentioned as such in the introduction section of the Wikipedia article on distributed lag [0].

Imagine a second order autoregression model for a time series that is a flat constant value for all time. How would you assign the coefficient for the first lag term and the second lag term? 0.5 to both? 1.0 to one and 0.0 to the other? In the presence of perfect correlation like that, it would be indeterminate how to assign effect sizes. Prediction accuracy of the model might be fine, but any causal interpretation of the effect sizes would be nonsense.

This also happens greatly for lagged regressors for more standard regression problems, and you often have to do something like z-scoring them and then combining the correlated regressions together into some type of pre-treatment aggregate score, like an average... effectively reducing N correlated lag terms into 1 composite score, when the correlations are high enough to cause multicollinearity problems. [1] has some further details.

In the context of pure forecast accuracy, you may not care so much about multicollinearity as long as the overall prediction is highly accurate.

But since this comment thread was about causality, I think it's important to point it out here. It can cause problems for causal inference in many classes of models, and it is a reason why you have to do careful pre-treatments, instrumental variable methods, etc., in some cases.

[0] https://en.wikipedia.org/wiki/Distributed_lag > [1] https://stats.stackexchange.com/a/278049/8927 >

Re: The Marshmallow Test: What Does It Really Measure?

#276

Post by xapata »

mlthoughts2018 wrote: post #275
xapata wrote: post #274

Earlier quoted context omitted.

If multicollinearity were such an issue, you'd be rejecting a wide range of techniques, such as including a lag term as a regressor for autoregressive models. In fact, you could consider this situation as an autoregressive process, using generations instead of traditional time steps.

Multicollinearity is a huge issue when fitting autoregressive and distributed lag models. In fact, it is even mentioned as such in the introduction section of the Wikipedia article on distributed lag [0]. Imagine a second order autoregression model for a time series that is a flat constant value for all time. How would you assign the coefficient for the first lag term and the second lag term? 0.5 to both? 1.0 to one…

Those are good points. Your initial comment criticized the endogeneity issue, not the multicollinearity.

Further, controlling for parents income is essentially an AR1 model, which doesn't suffer from the multicollinearity interpretation issues you described. The problem would be if child willpower were too highly correlated with parents' income, not if parents' income were highly correlated with child's income.

Re: The Marshmallow Test: What Does It Really Measure?

#277

Post by mortehu »

chronial wrote: post #267
mortehu wrote: post #169

Earlier quoted context omitted.

If the relationship is `y=x₀+x₁+𝓝`, then x₀ is more strongly correlated to `y-x₁` than it is to `y` alone. This holds even if x₀ and x₁ are strongly correlated, as long as they're not 100% correlated.

If y = x_0 and x_1 = x_0 + N, then y = x_1 - N. Then y and x_1 are highly correlated. But y - x_0 = 0 – which is not correlated to x_1 at all. y is "success" x_0 is "socioeconomic background" x_1 is "result of the marshmallow test"

The question isn't whether you can produce absurd example models where "What [vanderZwan says] would only be true if character traits and socioeconomic background were uncorrelated", it's whether it holds in general.

Re: The Marshmallow Test: What Does It Really Measure?

#278

Post by reitanqild »

thegabez wrote: post #182
reitanqild wrote: post #179

Earlier quoted context omitted.

Because IQ is a very specific thing. It doesn't mean you know a lot or are good at high school maths. IIRC an ideal IQ test would be something you could take even if you can't read and haven't learned maths and it will tell you something about your general problem solving skills. And as we all know, people can be excellent problem solvers without being very athletic at all, so mixing that into the score would actuall…

You don't need to know how to read or know math to shoot a basketball and the problem is well defined. Put ball through hoop.

Goes both ways: you don't need to be good at sports to be an excellent problem solver in the office.

Re: The Marshmallow Test: What Does It Really Measure?

#279

Post by TomMckenny »

haihaibye wrote: post #250
TomMckenny wrote: post #242

Earlier quoted context omitted.

This very article shows that rich tend to succeed and poor tend to fail regardless of whether the grab the marshmallow or not. But a summary of other issues include: If intelligence corresponds to wealth, we would expect Nobel prize winners and widely cited scientist to be the richest people. Royal families and their descendants, having concentrated their genes, should also be the most inventive, artistic or scientif…

>> rich tend to succeed and poor tend to fail regardless of whether the grab the marshmallow or not. We were talking about intelligence. The marshmallow test measures delayed gratification. >> If intelligence corresponds to wealth, we would expect Nobel prize winners and widely cited scientist to be the richest people. It's a correlation - so we'd expect them to be richer than average, and richer on average than dumb…

>and richer on average than dumber scientists

...but not the richest of all people. Nor do the richest people in the world tend to be great scientist, mathematicians, chess players, writers etc.

Because intelligence can only shuffle you around in the class you inherit but only luck will get you out of it.

Children born to carpenters will tend to become good carpenters because they get connections, learn skill and are given tools by their parents not because they are inheriting more intelligence than someone just breaking in to carpentry.

Children born to massive landlords will be landlords because they inherit the estate not because they are smarter than people wanting to become landlords with no cash.

>Wouldn't controlling for inherited wealth completely wipe away the signal?

You need to measure delta in wealth so in fact you _must_ control for wealth. Otherwise a rich kid who looses half his inheritance will be counted more successful than a poor kid who doubles his.

>just find a richer ethnic group

I point out above why this is impossible. If, for some reason, you decide this moment in history was the one to make the judgment on, you would find that the richest countries are also the most progressive and have the largest social safety nets. Are we going to claim the intelligence correlates to progressive politics and the wisdom of social safety nets as much as wealth?

>The marshmallow test measures delayed gratification.

And yet when it was believed to predict future wealth, it was said delayed gratification _was_ intelligence.

>the original marshmallow experiment showed...

The article explains why the original test was fundamentally flawed. For one reason, because it only had 90 subject rather than 900

Re: The Marshmallow Test: What Does It Really Measure?

#280

Post by ItsMe000001 »

Retric wrote: post #271
ItsMe000001 wrote: post #266

Earlier quoted context omitted.

> I think you are failing to grasp the price differences and my argument. Thank you very much! I think I DO understand your argument, but I have the suspicion that you fail at understanding my very first reply already, which sums it up quite nicely. Your "argument" is none because it addresses an issue that does not exist. You countered something that you yourself invented, an issue that does not exist. Please, just…

Yea, I read what you said as they don't need to eat that many calories, but over time their body says they do. Do you know any poor people? People that grew up in poor families? Just like everyone else when overweight their body wants to maintain that weight. That's going to take food, and in some cases a lot of it which is a real cost. Eating cheap calorie dense food can cover based nutrition requirements. It's not…

Di you know that you are famous?

http://dilbert.com/strip/2015-06-07

This fixation on the idea "people eat only vegetables!" is a fabrication of YOUR mind. You keep bringing it up. Are you sick? Do you need professional help? I'm a little concerned for you my little stupid friend. You can't tell what's only in your head from what what's outside? How about you stick to what I wrote and stop posting utter nonsense and made-up bullshit?

Post reply on HN