TALEB USES MATHEMATICS TO BUST THE MATHETMATICS OF HEREDITY USED BY RACEMONGERS

[Crusile]

@Ozell Israelite Thoughts? TLDR: White racists separate black twins at birth to "prove racism" then their mathetmatics get busted by taleb's mathetmatics

 

[whiteissuperior]

@Ozell Israelite Thoughts? TLDR: White racists separate black twins at birth to "prove racism" then their mathetmatics get busted by taleb's mathetmatics

i have a question. do you actually understand the math in this video? I do, because this is my line of work. but do YOU OP?

 

[Crusile]

i have a question. do you actually understand the math in this video? I do, because this is my line of work. but do YOU OP?

No can you please TLDR the video.
I actually wrote "please someone tldr video" in the original OP but then I had to rewrite the thread

 

[whiteissuperior]

No can you please TLDR the video.
I actually wrote "please someone tldr video" in the original OP but then I had to rewrite the thread

only watched the first 5 minutes so far, but upon watching this, I immediately knew no one without a masters in stats or undergraduate degree in mathematics or at the very least CS would understand what he meant. So why are you sharing videos that you dont understand.

Basically, in the first few moments, correlational studies are done via linear regression and finding the correlation coefficient.What ppl dont understand about statistics is that it involves a lot of assumptions. For example, to do linear regression, we assume that all real world data was generated from a statistical model, let our measurement be y, and our data be X. We assume our data came from some value Y = B +Bx + W, where W is some Gaussian noise, which is just a normal random variable with some mean and some standard deviation. Given our data, we then try to find the value of B that minimizes the squared distance of our predict data, which will be a line, with our actual data, which will almost never be linear. There arre obvious flaws with linear regression. I.e., what if data is non linear? Statisticians will often log or square transform the data to make it fit linear models, but thats aside from the point. It's relatively arbitrary, and why statistics is an art based on a LOT of assumptions. Anyway, correlation coefficients are a bit different that calculatiing linear regression coefficients, but the same assumptions hold. So what if the data is some piecewise function? As in, the relationship between IQ and success is linear for some populations, but for others it is not. Based on our statistical assumptions, when runjning linear regression, we are assuming all data is comes from a linear model with added noises, so if it is the case that our data comes from a piece wise function, it will appeal to be highly correlated. Additionally, if A is correlated to B, and B is correlated to C, it does not mean A and C will be strongly correlated. SO something like using SAT as a proxy for IQ, as these studies often do, is mathematically very unsound.

Finally, the last part I watched, up to the 4 minute mark, essentially says the method theyre using to calculate difference between correlations of identical twins and fraternal twins may be "non convex". Basically, if you've taken calculus, a strictly convex function is one such that there is only one minimum in the function, and thus the minimum value can often be found with 100% guarantee. Finding the optimal, minimal solution to a convex function is the foundation of many/most ML and statistical methods. You have a loss function (i.e. squared difference between predicted and actual value), and you seek to minimize it. The values that minimize it will bne the output to ur solution. However, if something is nonconvex, and does not have a closed equational form, as most real world data doesnt, you will often have to run a method called "gradient descent" to find the minimum/optimal value. And gradient descent has absolutely NO guarantee to converge at the minimum.

Social science is bunk, because social scientists dont really have the knowledge and understanding to analyze the findings of their own reseatch in a mathematically sound way. ALl they know is plug and chug data into a formula and output a p value, regression coeffecients, or correlational coefficient. So they dont understand the assumptions I just listed, and therefore do not know how to interpret their results. This is why stats is an art just as much as a science.

@smvmaxxertilllate you think if JS sees my based explanation of stats theyd hire me??

 

[whiteissuperior]

completely unironically, understanding statistics makes it hard to believe in modern day science. All it is is parameter fitting models, which works fine for biology, computer science ...etc. But its questionable how much further this will take us in physics (last great physics discovery, Higgs Bozon, was literally discovered just using ML/parameter estimation techniques and calculating probabilities), and in psychology/social science its definitely cope.

A year ago I would have agreed with race science, but the more I am educated, the less I agree with race science, not because the liberals brain washed me, I hate the liberals, but because I understand the assumptions behind statistics, and how the assumption that real world data can be reduced to linear, quadratic, cubic, or logarithmic functions is likely false. In real life, it probably is more of a piece wise function

 

[yougottadowhatyougot]

racemongers, when will they learn

 

[whiteissuperior]

Watched the rest cuz I actually. had fun explaining.

BAsically, conditional probability is just the probability of an event A happening GIVEN we know event B happened. We write this as P(A|B). And just the probability event A happend is P(A). If A is independent of B, (i.e., A or B happening have no effect on the likelihood the other happens), we say they are independent, and thus P(A|B) = P(A). However, this does not hold up for real world data. As I said above, statistics is based on assumptions of probabilities, so when taking the data of IQ versus work, the correlation between the environment and IQ can confound true results. Let us simplify this by calling A the event I succeed in life, and let's called B the event that I grew up in a good environment, and C the event I grew up in a bad one.

If I only correlated P(A|C) with IQ, or P(A|B) with iq, mathematically, I am not calculating the true correlation of IQ with success. I am calculating it conditioned on growing up in said environment. However, if I try to separate out the environment, some another assumption for regeression/correlation to work is homoschedacity, which I wont go into, but it basically means the noise is the same everywhere. IF this is not true, it also gives bad results.

tldr, mathematically, when trying to correlate IQ with success, it doesn't hold. Social scientists who say "x trait is x% hereditary" dont undersand the assumptions behind their models, and truth be told, no mathematical model can perfectly quantify real life. And one that could would likely be a neural net with like a million layers that you wouldn't be able to understand anyway

 

[whiteissuperior]

Did you actually understand anything I just said, or no? @Crusile

 

[mogger123]

completely unironically, understanding statistics makes it hard to believe in modern day science. All it is is parameter fitting models, which works fine for biology, computer science ...etc. But its questionable how much further this will take us in physics (last great physics discovery, Higgs Bozon, was literally discovered just using ML/parameter estimation techniques and calculating probabilities), and in psychology/social science its definitely cope.

A year ago I would have agreed with race science, but the more I am educated, the less I agree with race science, not because the liberals brain washed me, I hate the liberals, but because I understand the assumptions behind statistics, and how the assumption that real world data can be reduced to linear, quadratic, cubic, or logarithmic functions is likely false. In real life, it probably is more of a piece wise function

don't know enough about physics to comment on what you've said here but you're definitely right about social sciences. 99% of it is just forcing linear regressions where they dont belong or hypothesis testing again and again with no regard for assumptions until they get a result they like the look of

 

[smvmaxxertilllate]

only watched the first 5 minutes so far, but upon watching this, I immediately knew no one without a masters in stats or undergraduate degree in mathematics or at the very least CS would understand what he meant. So why are you sharing videos that you dont understand.

Basically, in the first few moments, correlational studies are done via linear regression and finding the correlation coefficient.What ppl dont understand about statistics is that it involves a lot of assumptions. For example, to do linear regression, we assume that all real world data was generated from a statistical model, let our measurement be y, and our data be X. We assume our data came from some value Y = B +Bx + W, where W is some Gaussian noise, which is just a normal random variable with some mean and some standard deviation. Given our data, we then try to find the value of B that minimizes the squared distance of our predict data, which will be a line, with our actual data, which will almost never be linear. There arre obvious flaws with linear regression. I.e., what if data is non linear? Statisticians will often log or square transform the data to make it fit linear models, but thats aside from the point. It's relatively arbitrary, and why statistics is an art based on a LOT of assumptions. Anyway, correlation coefficients are a bit different that calculatiing linear regression coefficients, but the same assumptions hold. So what if the data is some piecewise function? As in, the relationship between IQ and success is linear for some populations, but for others it is not. Based on our statistical assumptions, when runjning linear regression, we are assuming all data is comes from a linear model with added noises, so if it is the case that our data comes from a piece wise function, it will appeal to be highly correlated. Additionally, if A is correlated to B, and B is correlated to C, it does not mean A and C will be strongly correlated. SO something like using SAT as a proxy for IQ, as these studies often do, is mathematically very unsound.

Finally, the last part I watched, up to the 4 minute mark, essentially says the method theyre using to calculate difference between correlations of identical twins and fraternal twins may be "non convex". Basically, if you've taken calculus, a strictly convex function is one such that there is only one minimum in the function, and thus the minimum value can often be found with 100% guarantee. Finding the optimal, minimal solution to a convex function is the foundation of many/most ML and statistical methods. You have a loss function (i.e. squared difference between predicted and actual value), and you seek to minimize it. The values that minimize it will bne the output to ur solution. However, if something is nonconvex, and does not have a closed equational form, as most real world data doesnt, you will often have to run a method called "gradient descent" to find the minimum/optimal value. And gradient descent has absolutely NO guarantee to converge at the minimum.

Social science is bunk, because social scientists dont really have the knowledge and understanding to analyze the findings of their own reseatch in a mathematically sound way. ALl they know is plug and chug data into a formula and output a p value, regression coeffecients, or correlational coefficient. So they dont understand the assumptions I just listed, and therefore do not know how to interpret their results. This is why stats is an art just as much as a science.

@smvmaxxertilllate you think if JS sees my based explanation of stats theyd hire me??

Maybe

 

[Deleted member 25438]

I would like to reply to this but unfortunately I've posted my face here so if I get doxxed I'd get my life ruined

 

[Xangsane]

@Ozell Israelite

 

[Crusile]

I would like to reply to this but unfortunately I've posted my face here so if I get doxxed I'd get my life ruined

Nice excuse but ik you dont know the mathematics. Post the mathetmatics that debunk taleb's mathamatics or shutup

Did you actually understand anything I just said, or no? @Crusile

Good explanation

 

[Deleted member 25438]

Nice excuse but ik you dont know the mathematics. Post the mathetmatics that debunk taleb's mathamatics or shutup


Good explanation

Pm me ur discord