Editor's note:
The power of science is very powerful, but the truth of science is not far away, it is contained in the market. Listen to Zhang Zhongyuan, professor of the School of Statistics and Mathematics of Central University of Finance and Economics, to talk about the common statistical traps in life.
Zhang Zhongyuan's speech video:
The following is the transcript of Zhang Zhongyuan's speech:
When you say "statistics," you might think of CPI or GDP. This is the content of government statistics and belongs to an important branch of statistics.
In fact, the impact of statistics on our daily lives is far deeper and broader than these numbers.
For example, the drugs we all take seem to be the work of the pharmaceutical industry, but statistics contribute a lot to it. Without statistical methods, we have no way of knowing if the drug is a placebo. What is a placebo? A starch tablet, the doctor said that it is pain-relieving, as a result, 1/4 of the patients who eat starch tablets will report that it is no pain or the symptoms are reduced, in fact, he is only eating starch tablets - the power of the placebo effect can be seen. If we remove the placebo effect without statistical methods, the drugs sold on the market may not really be curable.
In fact, many of the major decisions you make in your daily life are made by observing many things around you and consulting many people around you. This process of "collecting information - removing the false and preserving the truth - and finally drawing conclusions" can be called "statistics".
Speaker Zhang Zhongyuan: "Scumbags are popular with girls? Beware of statistical traps
Statistical methods are very closely related to our daily lives, but they are not good by intuition. Statistics can easily be misused and abused.
For example, we often think of two things that are related to each other in daily life as causal, but in fact, this is often not the case, and this error is statistically called "correlation does not equal cause and effect".
In simple terms, the rooster is seen crowing and then the sun rises, but we cannot say that the sun is called by the rooster.
For example, big data analysis shows that when ice cream sales are high, the number of pool drowning accidents will also increase, so the two things of "pool drowning accidents" and "ice cream sales" are related - they have a common reason, that is, the season - but they obviously have no causal relationship. This is a relatively simple case.
When we encounter more complex situations, we tend to make mistakes. For example, seeing a child playing games every day, and then his test scores are relatively poor, parents will naturally link these two things, and then simply take "playing games" as the reason for his poor test scores - in fact, are these two things only related and not causal? Do they have some common cause? When parents think about this clearly, they can provide better education for their children.
There are more covert cases. For example, this is the picture I saw in the circle of friends. Say, Nobita and I are almost useless, why don't I have that jingle cat? In fact, in Nobita, "useless" and "with a jingle cat" are related, but you can't take them as cause and effect, so you are useless and don't necessarily have that jingle cat. (This is how my circle of friends is used to raise the bar, you can imagine how many friends I can have...) ）
Another example is this:
How did she heal? Everyone thinks to themselves.
There are also funnier examples.
The speaker made the mistake of thinking that certain film genres and specific consumer behaviors are relevant – but this is actually false correlation, that is, "correlation" itself does not exist, it is a coincidence. Why do you say that? Because "Fang Hua" was released in winter.
In addition, I want to emphasize that correlation is not equal to cause and effect, but correlation may contain cause and effect, so scientists often need to do more work to study. Of course, this is an afterword.
"I am a scientist" has a section called "I Ask a Scientist." Mr. Tong Dawei, the star who played the scumbag in the movie, asked this question: Why are scumbags still liked?
I may say that this is a pseudo-question. This is not to ask me, ask me, this program will end in one episode.
(With such a circle of friends, you can imagine that I really don't have many friends anymore...) ）
Why is it a pseudo-question? Let me take my time.
In my opinion, to clarify this issue, it is necessary to divide it into at least three levels.
The first level, what is "slag"? I understand it to mean "flower heart".
On the second level, does the girl know him scum? If the girl doesn't know his scum, it's called a scam and it's not in our discussion.
The third level, the girl knows that he is scum, and under the same conditions, he does not like him more than scum, that is, if the boy is not scum, does not care, the girl does not like him so much - does this thing exist? I had a big question mark in my heart.
Speaker Zhang Zhongyuan: "Scumbags are popular with girls? Beware of statistical traps
Why do people have this perception? There have also been studies in statistics, involving at least two knowledge points, one is survival bias and the other is implied variables.
Survival bias means that everyone will only pay attention to or pay more attention to those surviving and successful cases, and will ignore those failed and unsurviving cases, and then analyze what experience and characteristics those successful cases that survived, which is actually of little value.
Specific to this problem, you only pay attention to or see those scumbags who are loved, and ignore the large number of scumbags that no one loves, and then analyze what are the characteristics of these loved scumbags, is this thing valuable? Success learns to follow this routine, only analyze what are the characteristics of those successful cases, but if you take it seriously, it will be wasted.
For example, this is once a year: there is a feature of the college entrance examination champion, do your children have it? What are the features? Is it "all say Chinese"? I am so happy in my heart, and my children also have it.
The second problem is implicit variables. Of course, some scumbags survived and someone loved, but the attributes of this "scum" are a minus for him, and the reason why he has someone to love may be his other attributes. Therefore, if you ignore those other attributes here, it is not reliable to simply analyze the matter of "scumbags are loved".
Of course, we admit that in the eyes of different girls, the importance of these attributes is not the same. For example, some people are face control, value appearance, as long as there is a face, other things are not important; There is an uncle control, as long as it is mature, everything else is not important; Others value the economic situation of boys, but I really haven't heard of slag control.
Speaking of implied variables, let's look at the following study.
Such a paper, published in a journal in the United States, said that "cleaning products damage women's lungs equivalent to smoking a pack of cigarettes a day, but have no effect on men's lungs." Seeing this, I almost didn't cry, what kind of toxic research is this, in order to let boys do housework so hard? Isn't it weird that "it has such a big impact on girls, but no effect on boys' lungs"? As long as it comes to research such as "boys do housework", I will take it seriously and must write an article to refute it. (After all, books cannot be read in vain.) Read more books and you will be able to be invincible in the family. So after my divorce □□□□)
Okay, let's get back to business.
This study, if I want to look at it, ignores an implicit variable: cooking. Women who love cleanliness are often more willing to cook, and cooking hurts the lungs, while for men, cleaning is more likely to be a profession for him, and it has nothing to do with whether he cooks or not.
Let's look at this again. If we consider the mortality rate of oncology departments in different hospitals, you will find that the mortality rate of oncology departments of famous hospitals may be higher than that of lower-level hospitals. Does that mean that the level of doctors in famous hospitals is not very high? This is certainly not true. Because an implicit variable is also ignored here: tumor type. Most of the visits to top hospitals are refractory tumors, and the mortality rate itself is high. But this situation is more obvious.
The following example is not so obvious, and you need to think a little.
People found that the acceptance rate of female students in the admission of Berkeley University in the United States is significantly low, so they questioned the problem of "gender discrimination" in Berkeley University - this thing seems to be fine, the data is here, what else to say? But that's your reaction. At Berkeley University, a hair is divided into 8 petals, which is very thin, and it is not easy to say that there is a problem. Of course, this is not the case.
Because, if you look at it by major, you will find that the acceptance rate of female students in each major is actually higher. The result of the "low overall acceptance rate for girls" is due to the fact that the majority of girls apply to colleges that are inherently difficult to admit, while the majority of male students apply to colleges that are themselves high acceptance rates. This phenomenon is called the "Simpson paradox."
Let's look at the following example.
This is 2 groups of 10 people each. Everyone plays a game of play, and the group with the most people wins.
There are two types of levels, one is easy mode and the other is hard mode. No matter which level you choose, as long as you beat it, you'll win.
Then for various reasons, the first group preferred to choose the simple mode. There were 10 people in total, and 8 of them chose the simple mode. The second group preferred Hardmode, with 10 people in total, and 8 of them chose Hardmode.
The results of the competition are as follows: in easy mode, 7 people from the first group have passed the level and won; The second group all passed. So in simple mode, the second group of people has a higher pass rate. Similar to the analysis, in Hardmode, the second group of people also had a higher pass rate. But overall, the pass rate of the first group is higher. This phenomenon is the "Simpson paradox", because the first group of people choose the simple mode more, and the simple mode itself has a higher pass rate. Going back to the example of "cancer hospitals and mortality". The doctors in those famous cancer hospitals are like those who prefer to choose the hard mode, which is a misery in itself.
So much misuse of statistics has just been talked about, in fact, we must also pay attention to the fact that statistics cannot be abused.
If we abuse statistics in social governance, the consequences will be serious. For example, if we live in Beijing, if we count the crime rate of people from various provinces, if, I mean if the statistics conclude that the crime rate of northeasterners is higher, then can we ask northeasterners not to go out at night? Obviously this cannot be done.
Why can't this be done? Because we apply a group-based overall conclusion to individuals, there may be other implicit factors to this conclusion. And everyone is precious, and even if this kind of decision hurts an individual, the cost is huge and irreparable. When we make decisions, we balance losses and gains, and this kind of decision-making is huge, unethical, and is a statistical abuse. It has a technical term called "statistical discrimination", and this kind of decision-making does not happen. However, statistical discrimination often occurs in everyday life. What about wearing colored glasses and treating others differently according to gender, race, and place of origin?
However, then again, you can imagine a scene: a girl returns at night, and there is a boy behind you who has been following you - at this time you should run or run, you can't say, "Teacher Zhang said, if I run, I treat him as a bad person, this is statistical discrimination, I have to exchange a WeChat chat with him first to see if he is a bad person." In this way, it is rigid, there are dates without dates to hit three shots, this is not okay.
Remember: you should use statistics with caution and not abuse them.
Finally, as a statistician, I want to make a statistic: ask friends who did not come to raise their hands. No one raised their hand, it seems that all the popular science enthusiasts in Beijing have come, I am asking for fire. Thank you for your patience.
Speaker Zhang Zhongyuan: "Scumbags are popular with girls? Beware of statistical traps
Author: Zhang Zhongyuan
Editor: Malt Poplar