10 statistical methods commonly used for "dry goods" data analysis, with key application scenarios attached

Estimated reading time ：6min

Reading suggestions ： This article is a summary of dry goods , Long term use , Suggest collecting before reading .

Solve the pain ： Recently, I received some private letters from my classmates , One of the most frequently asked questions is ：“ What statistical knowledge do you need to learn to do data analysis ？” Therefore, I would like to share some statistical methods that are widely used in my work .

00

preface

From the perspective of origin , Data analysis is 「 statistical 」 And 「 Computer 」 Interdisciplinary , Statistical knowledge is applied ; From a work point of view , Apply statistical knowledge , It can measure the value of data to the business more scientifically .

therefore , As a data analysis student , It is necessary to master basic statistical knowledge . below , Little fire dragon summarized 「 Commonly used in work 10 Two statistical methods 」.

01

Descriptive statistics 「 Commonly used index ：5 star 」

Descriptive statistics , Through the general mathematical method and chart method , Describe the business data and its distribution status , It is most commonly used in work .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• weekly / Monthly report ：「 If you complete a high-quality data weekly report / Monthly report 」

02

Hypothesis testing 「 Commonly used index ：5 star 」

Hypothesis testing , Used to judge samples and samples 、 The difference between the sample and the population , It is caused by sampling error , Or because there are differences in itself . It mainly covers 「 Parameter test 」 and 「 Nonparametric test 」, The concepts of both are as follows ：

Parameter test ： Suppose the data obey a certain distribution （ It is generally normal distribution ）, The population parameters are tested by the estimators of sample parameters .

Nonparametric test ： Regardless of the overall distribution , Test the distribution of data directly .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• Transaction analysis ：「 Three ways to quickly locate abnormal dimensions 」
• Causal analysis ：「AB Best process of experiment 」

03

Contingency table analysis 「 Commonly used index ：3 star 」

Contingency table analysis , It is used to judge whether there is obvious correlation between discrete variables . for example ： Whether there is correlation between performance grade and gender .

The contents are as follows ：

The main application scenarios are as follows ：

04

Correlation analysis 「 Commonly used index ：4 star 」

Correlation analysis , It is used to judge a certain relationship and degree of correlation between phenomena , for example ： positive correlation 、 negative correlation , It is frequently used in exploratory analysis .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• User growth ：「 My understanding of user growth 『 New users 』」

05

variance analysis 「 Commonly used index ：2 star 」

variance analysis （ also called F test ）, A significance test used to measure the difference between the mean values of two or more samples .

The contents are as follows ：

06

regression analysis 「 Commonly used index ：5 star 」

regression analysis , Used for fitting daily indicators , And predictions of future trends , It is widely used in work .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• Index prediction ：「 Time series prediction artifact -Prophet『 Theory Chapter 』」、「 Time series prediction artifact -Prophet『 Realization chapter 』」

07

Clustering analysis 「 Commonly used index ：4 star 」

Clustering analysis , Used to transfer the user to / Content , Without a priori guidance , To divide into categories .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• Clustering analysis ：「 On the application of cluster analysis in work 」

08

discriminant analysis 「 Commonly used index ：4 star 」

discriminant analysis （ Classification problem ）, Judge the category according to the characteristics of the research object . The difference from cluster analysis lies in , Cluster analysis before analysis , I don't know how many categories there are and what they are , And discriminant analysis is when the total category is known , Judge the group to which the new sample belongs .

The contents are as follows ：

The main application scenarios are as follows ：

【 Application scenario reference article , Can stamp the blue part 】

• User growth ：「 My understanding of user growth 『 Loss warning 』」

09

Principal component analysis 「 Commonly used index ：2 star 」

Principal component analysis （Principal Component Analysis,PCA）, Is a set of possible correlation variables , Into a set of linearly unrelated variables , The transformed set of variables is called the principal component .

The most important function of principal component analysis is 「 Dimension reduction 」, It can also be used for 「 Explore the relationship between variables 」. A brief explanation , In the process of modeling , Many variables are often selected as features , And there is often a correlation between these variables , This will cause 「 Multicollinearity problem 」. therefore , Need a way , Transform these variables into relatively independent and as much information as possible covering the original variables , The principal component is one of them , Transform the original variable into several new variables that are independent of each other .

Eliminate illiteracy - Multicollinearity problem

The independent variables （ features ） Because there is a correlation between , Thus, the model estimation is distorted （ The result is not stable , for example ： Contribution degree of random forest characteristics , The results of multiple runs are quite different ）.

10

Factor analysis 「 Commonly used index ：2 star 」

The function of factor analysis is the same as that of principal component analysis , Also for 「 Dimension reduction 」. The principle is between multiple independent variables , Look for potential factors , Similar variables are grouped into a factor , Replace the original independent variable with a factor .

The same as principal component analysis ： Play the role of cleaning up the internal relations in the original independent variables .

Different from principal component analysis ： Principal component analysis focuses on the information of aggregated variables , Factor analysis focuses on the information of explanatory variables , Principal component analysis is a subset of factor analysis .

above 10 Three statistical methods are common in work , But the way is more than that , It also includes ： Reliability analysis 、 Survival analysis 、 Multiple response analysis 、 Distance analysis, etc . Waiting for you to continue to explore in your work 、 Explore the scene , And apply the knowledge .

The above is the content sharing of this issue .