How far is the first kind of mistake from us

I haven't blogged for a long time , If you want to send it, you don't know which one is better . Then whatever .

quote 【 Necessary skills for data analysis ｜ Correctly interpret your AB experiment 】https://cloud.tencent.com/developer/article/1618910

Thinking of data analysis ： 

When you want to analyze a thing and draw a conclusion , You have to look for indicators . When this is not data analysis thinking , Data analysis thinking should be thinking about how to prove the reliability of indicators , for instance , I want to prove that eating chocolate affects people's mood , It has to be made for chocolate ab test , But I have to prove the correlation between chocolate and the experimental results , namely

【 How to measure whether there is an indicator of significant changes 】

Next I have to analyze the overall standard deviation of people's mood fluctuations （ This value should be able to check the table , If you can't check the table, estimate it , It can be estimated according to the relevant knowledge of sample distribution ）, Then study how many overall standard deviations the average value of mood swings of people who ate chocolate deviated , If so x A standard deviation , Then check the table x Probability distribution of standard deviation .

Key concepts ：

Overall standard deviation , The difference between the average value of the experimental group and the control group , Probability distribution of multiple of standard deviation .

Popular said , Is to use probability distribution （ normal distribution ） To measure whether the band meets the normal range . The standard deviation can be regarded as the unit to measure the band .

With this probability distribution , We can know , Whether the mood fluctuation of people who eat chocolate is a normal phenomenon , For example, we 【 Think 】80% People will have such fluctuations , Then it's a normal phenomenon , The conclusion is that chocolate has no effect on mood .

But the problem is what we think 【80%】 Is this standard too much , Or the requirements are too low ？

Here comes a concept ： The first kind of error and the second kind of error .

The first kind of mistakes are mostly misplaced , Because people are inherently insecure , So always ask too much , For example, we will require that fluctuations must 99% All people have , Most people will have , It is normal fluctuation , This is also an overly pessimistic approach .

let me put it another way , People are naturally pessimistic , Depressive tendencies .

It is easy to make the first kind of mistakes , Cause rejection of the truth , Or say , It's easy to miss beautiful things .

On the contrary, people who are overly optimistic , But it is easy to make the second kind of mistakes , It leads to the wrong belief in villains .

How to avoid the first kind of mistakes , Namely , Access to information , Make yourself feel safe .

The way to avoid the second kind of mistake is , Improve your screening ability , For example, improving the bottom line of communication can also .