Popular explanation: what is a clinical prediction model

获取更多R语言知识,请关注公众号：医学和生信笔记

“

医学和生信笔记.专注RThe use of language in clinical medicine,RLanguage data analysis and visualization.主要分享RLanguage to do medical statistics、meta分析、网络药理学、临床预测模型、机器学习、生物信息学等.

很多人,The concept of clinical prediction model is very blur,A: you get to the Internet,Are all proper nouns,And then tell you has been in clinical use for a long time,很有用,All kinds of magazines that the cow force can see the figure of clinical prediction model,Born in letter article is common.

一通操作下来,你迷茫了,Clinical prediction model is more don't know what,感觉懂了,And feel not understand！

Today in plain language to explain what is a clinical prediction model,读完后,You should really know！

• 简单理解
• Clinical prediction model and machine learning
• Clinical prediction model and statistical

简单理解

临床预测模型,初听这个名字,May feel very tall,其实没那么复杂,你Can understand as a way to,This method can predict whether a person is sick or have no disease,Or predict a person after a period of time will not die,Or predict a person's a disease will recur,Or the prediction is a sample of the tumor and normal tissue......

再直白一点,Clinical prediction model is一个公式,根据这个公式,You to provide some basic information,比如年龄、性别、体重、Quantity of hemoglobin, etc（Or a gene expression quantity etc）,This person can be calculated exactly is sick or not！

At present many diseases do magnetic resonance、做CT、Pathology to diagnose,If you have found a formula,As long as a blood test,Get several biochemical index,Can according to your formula to calculate out the person is sick or have no disease！这不比CT、磁共振、Pathological easier？值得推广.

所以,The essence of the clinical prediction model is a kind of分类方法.通过这种方法,You can categorize a lot of things,比如,生和死、Sick and have no disease、肿瘤和非肿瘤、Recurrence and no recurrence, etc.

Since it is a way,It must have accurate and inaccurate,看名字也能知道,这只是一种预测,或者叫：猜（According to the guess）！ If you can this method and金标准相提并论,That means your method is very cow,If you just method is more simple and convenient、经济适用,That your method is really too much,Very wants to become a new gold standard！

That how to evaluate your method is good or bad？这就是Evaluation of clinical prediction model,Through various index（后面会详细介绍）、From the Angle of the various evaluation.

说了这么多,How can I get my model（或者叫方法）呢？This is another major contents of clinical prediction model：To build up the forecast model on the clinical.

前面说过,Clinical prediction model is essentially a formula！说个最简单的,逻辑回归（logistic）,Everybody should know how to construct logistic regression？Is the independent variable and dependent variable.Give you a few independent variables,A binary classification of dependent variable,大家通过SPSS点点点,You can get each independent variable coefficient,Then I can write a logistic regression equation.You of the logistic regression equation,This is a clinical prediction model！Give you a few independent variable values,根据这个方程,Can you work out the value of the dependent variable,Then you can classify！

说到这里,Believe that you should know that a lot！但是这还不够,You may have heard what machine learning、lasso、随机森林、支持向量机等等,别慌,这就是我们接下来要说的：Clinical prediction model and machine learning relationship.

Clinical prediction model and machine learning

机器学习,If sound is tall,But for make the clinical,Don't need to know too thoroughly,Probably know what is enough.

Logistic regression is a kind of machine learning,随机森林、决策树、支持向量机、lasso、岭回归、弹性网络、xgboost等等,These are like logistic regression,就是不同的方法而已！

Studied medical statistics all know（Have not learned may also know that）,If the dependent variable is continuous variable,We use multiple linear regression,If the dependent variable is a binary classification variables,就用logistic回归（分类）.回归和分类,Just two main task of machine learning.很多方法,比如随机森林,既可以做回归,又可以做分类,And the accuracy is high,This is why people like to use other methods,Mainly in order to improve the accuracy of.

临床预测模型,Just one of application in the field of machine learning in medical,回归和分类,适用于各行各业,So you have all heard the machine learning in many areas the words.此外,还有深度学习、人工智能等等,These can be simple to understand more of the cow force method！

These different approaches have their own suitable scene,In the right scenario to get the best performance,How to make the model better,Then you have to learn some basic knowledge of the machine learning,这些东西在bilibili一搜一大堆,大家可以自行学习,But don't be too addicted yo！

But you search machine learning tutorial,The are recommend you吴恩达、Watermelon books or something,我不推荐,These things are not show people in the field of biological medicine,Look at these,May be listening to the heavens,After all, many medical students,Even high number are not learn！我比较推荐statquest,bStand also can search,This is a foreign biometric teach,His style is more suitable for us！

Clinical prediction model and statistical

We studied the medical statistics,In some ways and machine learning is a intersection.比如,逻辑回归、多元线性回归,Both statistical methods,And machine learning algorithms,这并不冲突,Like a person in different places have different identity.

When talking about clinical prediction model,We may be biased towards the machine learning a little more,Use a lot of ways, after all,From the machine learning field.

You may see in many letter article,Using a model did not test conditions in advance,直接就用了.But in medical statistics,Many methods are suitable conditions,Eligible to use.What is the right of？

其实不用纠结,Others can you can,多看文章,You can find all kinds of usage,But others still sentSCI,你也可以.如果非要说区别,It involves the prior school and posterior school these things,I don't really understand,如果你有兴趣,可以自己探索.If the article is to,Then don't make the same,Read a few high scoreSCI,Follow the inside of the train of thought to imitate吧！

读到这里,You should have a general understanding of clinical prediction model,Don't fog.

如果觉得这篇文章不错,欢迎点赞、在看、转发,让更多人看到;如果你还有疑问,Also welcome to join the following communication group of,继续探讨！

获取更多R语言知识,请关注公众号：医学和生信笔记

“

医学和生信笔记.专注RThe use of language in clinical medicine,RLanguage data analysis and visualization.主要分享RLanguage to do medical statistics、meta分析、网络药理学、临床预测模型、机器学习、生物信息学等.