What's new

Body count data

Just an idiot that doesn't know shit about epidemiology.

All he did was graph some numbers Excel and picked a quadratic model. There's no reason to pick a quadratic model in epidemiology. Disease growth is typically exponential. And when you do best fit, guess what: the parameters are forced to fit as best as possible. Any smooth function can be fitted by arbitrary number of polynomial if you pick enough polynomials and cherrypick the coefficients.

His model is y = Ax^2 + Bx + C. So why those particular numbers of A,B,C, and what do A,B,C mean? He doesn't know, nobody knows, because it's a shit and unphysical model.

From the Taylor expansion of y=Ae^Bx = A+ ABx + A(Bx)^2/2! + A(Bx)^3/3! +.... you can always pick A and B such that y = Ae^Bx ~ y = A + ABx + A(Bx)^2. You can always pick random numbers that make an exponential look quadratic.
That r2 value of 0.9995 is what was unsettling. You can probably fit at most 10 points for any quadratic process of natural origin so well, but not for 20 days straight?

Your knowledge of statistics seem to be more fresher than mine. What do you think?
 
.
央广军事今天 07:20 来自 Weibo.intl
#直击疫情防控第一线# 【向下吧小黄线!湖北以外新增确诊病例连续5天下降】累计治愈病例连续8天超过累计死亡病例,#新增治愈连续10天超新增死亡病例#!加油!#武汉战疫# (人民日报)

China National Radio Military
today at 07:20 from Weibo.intl
# 直击 疾疾 防控 第一 线 # [Go down the little yellow line! Newly diagnosed cases outside Hubei have fallen for 5 consecutive days] Cumulative cured cases exceed cumulative deaths for 8 consecutive days. Come on! # 武汉 战 MED # (People's Daily)

670172e1ly1gbqy3exs21g20k00psn0m.gif
Purple line - daily confirmed.
brown line - daily confirmed (Hubei Province)
Yellow line - daily confirmed (outside Hubei)

670172e1ly1gbqy3fgew0g20k00wegqr.gif

Top diagram.
Blue line - cumulative recovered.
Red line - cumulative death.

Bottom diagram.
Blue line - daily recovered.
Red line - daily death.
 
.
China reports 3,062 new confirmed cases of novel coronavirus infection, 97 new deaths
Source: Xinhua| 2020-02-10 08:48:48|Editor: huaxia

138769602_15812957279881n.jpg
Patients infected with the novel coronavirus at a makeshift hospital converted from an exhibition center in Wuhan, central China's Hubei Province, Feb. 5, 2020. (Xinhua/Xiong Qi)

Chinese authorities reported 3,062 new confirmed cases of novel coronavirus infection and 97 new deaths as the total confirmed cases on the Chinese mainland had reached 40,171 by the end of Sunday, and 908 people had died of the disease.

BEIJING, Feb. 10 (Xinhua) -- Chinese health authorities said Monday it received reports of 3,062 new confirmed cases of novel coronavirus infection and 97 deaths on Sunday from 31 provincial-level regions and the Xinjiang Production and Construction Corps.

Among the deaths, 91 were in Hubei Province, two in Anhui, as well as one in Heilongjiang, Jiangxi, Hainan and Gansu provinces respectively, according to China's National Health Commission.

Another 4,008 new suspected cases were reported Sunday, said the commission.

Also on Sunday, 296 patients became seriously ill, while 632 people were discharged from hospital after recovery.

The overall confirmed cases on the Chinese mainland had reached 40,171 by the end of Sunday, and 908 people had died of the disease.

The commission added that 6,484 patients remained in severe condition, and 23,589 people were suspected of being infected with the virus.

A total of 3,281 people had been discharged from hospital after recovery.

The commission said 399,487 close contacts had been traced, adding that among them, 29,307 were discharged from medical observation Sunday, with 187,518 others still under medical observation.

By the end of Sunday, 36 confirmed cases including one death had been reported in the Hong Kong Special Administrative Region (SAR), 10 confirmed cases in the Macao SAR, and 18 in Taiwan.

One patient in Macao and one in Taiwan have been discharged from hospital after recovery.
 
.
You know why he said that right?

He was accusing China of cooking the data. Because that is what you get if you put actual data supplied by Chinese government This is the graph looks like

View attachment 604401

Most outbreak data are, as you said present in exponential increment, and also random, so you don't get a perfect quadratic graph if you plot an actual outbreak progress.

However, the graph above did not indicate an exponential increment and is quite conformal, which suggest the number is not random.

There are only one way the graph above, which I will remind you again is from official Chinese figure, that make sense, is that the outbreak was being contained nicely fit into a quadratic model (hence R-0 decrease quadratically) I mean, that usually not the case in outbreak, hence the data cooking rumour.

your so called "conformal" is horse shit. you can make any function fit any graph with r = 0.999999999999999 if you cherry pick your coefficients and don't have to physically justify it.

That r2 value of 0.9995 is what was unsettling. You can probably fit at most 10 points for any quadratic process of natural origin so well, but not for 20 days straight?

Your knowledge of statistics seem to be more fresher than mine. What do you think?

it's horse shit. you can make r = 0.99999 for any smooth function if you were willing to 1. fit enough polynomials and 2. cherry pick coefficients because you can expand any smooth function as a Taylor series.

they can't defend their use of a quadratic model. that is why when you have a model, it has to be physically defensible. a quadratic is not physical. it is like epicycles used when people thought the sun orbits the earth.

You plot those same numbers in Excel, you also have a logarithmic model fit with r = 0.997 and a power series model with r = 0.9997. You can even fit a cubic.

You can make any numbers you want fit.

also, why they use accumulated total? this is poor data choice, because some people can get cured. why don't they use time series of new admissions? because then it doesn't look like they can make a political point on Excel.
 
Last edited:
.
your so called "conformal" is horse shit. you can make any function fit any graph with r = 0.999999999999999 if you cherry pick your coefficients and don't have to physically justify it.

Yeah, everything is horse shit to you.

If you compare graphs between Ebola, SARS, Hanta and other virus, they aren't present in a Quadratic function.

Compare these 3 graphs

Virus.jpg

2019 Corona Virus

graph1-cumulative-reported-cases.jpg

2013/4 West Africa Ebola

worldwide2003_7_13.jpg


2003 SARS

Even a monkey can see the different. You argue Taylor Series is what mathematician do when they try to find a Polynomial (quadratic) function with a Non-polynomial function, but in this case, Chinese government is trying to achieve that. If you put R-0.0999999999999 you will indeed fit any graph into a Polynomial, but that is what we are talking about here. Real life case should not be, especially we are not talking about a big sample.

But hey, I forgot you are a Chinese member, when you call it horse shit, it must be horse shit, no explanation necessarily. And you don't need any data to support your argument, because you are a god of argument.
 
Last edited:
. .
Yeah, everything is horse shit to you.

If you compare graphs between Ebola, SARS, Hanta and other virus, they aren't present in a Quadratic function.

Compare these 3 graphs

View attachment 604409
2019 Corona Virus

View attachment 604410
2013/4 West Africa Ebola

View attachment 604411

2003 SARS

Even a monkey can see the different.

But hey, I forgot you are a Chinese member, when you call it horse shit, it must be horse shit, no explanation necessarily. And you don't need any data to support your argument, because you are a god of argument.

totally apples and oranges. I think my dog taught you math, because this sure as hell isn't human math.

Ebola can be approximated by a quadratic (and made to fit perfectly with cherry picked coefficients) if you look at only 10 days, which is what this data did. The data you showed is over several months. why doesn't this guy go back to January statistics? Because then it doesn't look the way he wants.

SARS isn't even an accumulated caseload but rather a time series of new cases. it is literally not the same measurement.
 
.
totally apples and oranges. I think my dog taught you math, because this sure as hell isn't human math.

Ebola can be approximated by a quadratic (and made to fit perfectly with cherry picked coefficients) if you look at only 10 days, which is what this data did. The data you showed is over several months. why doesn't this guy go back to January statistics? Because then it doesn't look the way he wants.

SARS isn't even an accumulated caseload but rather a time series of new cases. it is literally not the same measurement.

You didn't get the point, you arguing it the other way around, because in this case, you are accusing people put a Non-Polynomial into a polynomial equation, which is exactly what Chinese are doing, because actual live data should not be a Polynomial sequence.

You don't have an exponential event represent by a polynomial graph, especially with a smaller sample size like this.

But well, if your dogs teaches me maths, I guess your dog's poo is the one that taught you.
 
.
You didn't get the point, you arguing it the other way around, because in this case, you are accusing people put a Non-Polynomial into a polynomial equation, which is exactly what Chinese are doing, because actual live data should not be a Polynomial sequence.

You don't have an exponential event represent by a polynomial graph, especially with a smaller sample size like this.

But well, if your dogs teaches me maths, I guess your dog's poo is the one that taught you.

clearly you don't understand the simple concept of Taylor series. exponents can easily be represented by polynomials in a limited range, depending on how many polynomials you use.

I am not "accusing people put a Non-Polynomial into a polynomial equation", the **** this even mean? I don't understand gibberish. I'm saying that given any set of data, you can make it fit any equation you want with as high accuracy as you want, if you cherry pick the equation's coefficients. and if you can't understand that statement, then you have dog level math.
 
.
clearly you don't understand the simple concept of Taylor series. exponents can easily be represented by polynomials in a limited range, depending on how many polynomials you use.

I am not "accusing people put a Non-Polynomial into a polynomial equation", the **** this even mean? I don't understand gibberish. I'm saying that given any set of data, you can make it fit any equation you want with as high accuracy as you want, if you cherry pick the equation's coefficients. and if you can't understand that statement, then you have dog level math.

Exactly how stupid were you?

You are talking about Taylor Series, right? It is basically taking a non-polynomial function and finding a polynomial function. You always find a quadratic function within a non-polynomial function.

800px-Sintay_SVG.jpg


You are absolutely right, with the right constant, you can always find a Polynomial function within just about anything, but in this case, Virus Inflection is not a Polynomial Function, it was exponential function. And it would not fit on a perfect Quadratic graph due to low sampling structure (20 days) due to the standard deviation involved. It would be jumping all over the place. First of all, do you even know what is the probability to have a random event like virus outbreak to find the right R coeff and represent itself as a Quadratic when it should not be one? Second of all, how is it a total random event that can ever represent neatly? Yet,

However, according to the Chinese Data, the inflection rate are indeed quadratic, and if they are, you can basically predict the function, (Which the OP post do) mapping up with the official data. You need both situation above to be comply with each other to have that curve plotted using real live data, which in term of probability, its slimmer than I get hit by lotto in the next 3 draw.

Now, that is the reason why people accusing China for cooking data.

What you say, either you are very stupid and did not realise, or you are basically explaining how Chinese Cook the data by using Taylor Series. I mean even dog poop math will not be this stupid.
 
Last edited:
.
Exactly how stupid were you?

You are talking about Taylor Series, right? It is basically taking a non-polynomial function and finding a polynomial function. You always find a quadratic function within a non-polynomial function.

You are absolutely right, with the right constant, you can always find a Polynomial function within just about anything, but in this case, Virus Inflection is not a Polynomial Function, it was exponential function. And it would not fit on a perfect Quadratic graph due to low sampling structure (20 days) due to the standard deviation involved. It would be jumping all over the place.

However, according to the Chinese Data, the inflection rate are indeed quadratic, and if they are, you can basically predict the function, (Which the OP post do) mapping up with the official data.

Now, that is the reason why people accusing China for cooking data.

What you say, either you are very stupid and did not realise, or you are basically explaining how Chinese Cook the data by using Taylor Series. I mean even dog poop math will not be this stupid.

Taylor series means that you can approximate an exponential with a polynomial. In this case, because of lack of more than 10 data points, the quadratic approximation is close. you can make it fit any quadratic you want because there's only a few data points. if you changed the coefficients, there would be "spread". you can also change the coefficients so there is no spread. you can change the coefficients to make the error anything you want.

you clearly don't understand this simple concept. all you can do is pull wikipedia articles. Before I educated you on how to use Taylor series you didn't even know what that shit is, now you've been educated and think you can pull Wikipedia on me?
 
.
Taylor series means that you can approximate an exponential with a polynomial. In this case, because of lack of more than 10 data points, the quadratic approximation is close. you can make it fit any quadratic you want because there's only a few data points. if you changed the coefficients, there would be "spread". you can also change the coefficients so there is no spread. you can change the coefficients to make the error anything you want.

you clearly don't understand this simple concept. all you can do is pull wikipedia articles. Before I educated you on how to use Taylor series you didn't even know what that shit is, now you've been educated and think you can pull Wikipedia on me?

LACK OF DATA POINT?

There are 25 datum point, and while it is not that many, but that is not "lacking" of datum point.

And that graph is NOT an approximation from the actual datum as well, so coeff should not have changed, because it represent the natural occurrence, they are plotted with actual data supplied by National Health Ministry.

So, basically you are either telling me how China cook its data by using Taylor Series. or you don't know shit about graphical representation and statistical analysis.
 
.
LACK OF DATA POINT?

There are 25 datum point, and while it is not that many, but that is not "lacking" of datum point.

And that graph is NOT an approximation from the actual datum as well, so coeff should not have changed, because it represent the natural occurrence, they are plotted with actual data supplied by National Health Ministry.

So, basically you are either telling me how China cook its data by using Taylor Series. or you don't know shit about graphical representation and statistical analysis.

wheres the 25 data points on the original data used for the quadratic fit? it isn't there. count the data points for yourself:


I said nothing of the sort. you are presenting a false dichotomy. I am saying that anyone can fit any function to this data with high R factor if you cherry pick the coefficients, which was indeed what was done.

see, this is how you argue. You try to tire the other guy out by purposely misdirecting, by presenting false dichotomies, and basically, skirting on things you don't actually know about, and hoping the other guy gets frustrated. but see, that shit doesn't work when the other guy knows 100% what they're talking about. I'm not here to convince you. I'm here to educate the audience.
 
.
wheres the 25 data points on the original data used for the quadratic fit? it isn't there. count the data points for yourself:


I said nothing of the sort. you are presenting a false dichotomy. I am saying that anyone can fit any function to this data with high R factor if you cherry pick the coefficients, which was indeed what was done.

see, this is how you argue. You try to tire the other guy out by purposely misdirecting, by presenting false dichotomies, and basically, skirting on things you don't actually know about, and hoping the other guy gets frustrated. but see, that shit doesn't work when the other guy knows 100% what they're talking about. I'm not here to convince you. I'm here to educate the audience.

Oh my god. You are really "the word I cannot say otherwise I will get banned"

I am not talking about that reddit prediction curve, I am talking about the actual datam curve provided by National Health (Authority or whatever) of China

This is the curve

Virus.jpg


You do know that follow Benford's law if you mirror it (Which used to predict numerical distribution, also used to catch people faking data), right? And random event of natural occurance in term of probability would never follow Benford's Law. If you present this curve to a forensic accountant, he will tell you this is a fake graph. Yet, it was plotted with actual data from China

And LMFAO for saying I am the one who try to win an argument by making you frustrated. Because from where I am seeing, you are doing that.
 
Last edited:
. .

Country Latest Posts

Back
Top Bottom