Methods and prerequisites for various tests

(1) Normality Test: large samples are tested by K-S, small samples are tested by Shapiro-Wilk. There are two specific methods, one is to use descriptive statistics-> using E, one is to use non-Parametic test-> 1 sample K-S test (2) standardized processing (de-dimensional ): convert the original group of data into data that conforms to the N () Distribution to achieve de-unit effect. The specific method is to select Save standardized values as variables under descriptive statistics-> descriptive to obtain the corresponding standardized data. (3) One-Factor Variance Analysis: 1. prerequisite Normality Test, independence test, variance homogeneity 2. for a fixed effect model, data import can be implemented using compare means/One Way ANOVA or GLM/univariate. For a random effect model, GLM/univariate can be used. How to determine whether to use a fixed effect model or a random effect model: because the original hypothesis of Hausman test is: using a random effect model: Alternative Hypothesis: using a fixed effect model. therefore, simply look at the P value. If the P value is less than a significance level of 0.01/0.05/0.1 (depending on how you set the significance level), you can reject the original hypothesis, the fixed effect model is used. (4) multi-factor variance analysis 1. prerequisite Normality Test, independence test, variance homogeneity 2. data Import uses GLM/univariate for fixed and random effects. 3. model Selection for multi-factor variance analysis with repeated observed values, first analyzes whether there is an interaction effect between each factor. If there is no interaction effect, the interaction effect is considered as an error effect, only the independent effect or main effect of each factor is analyzed. (4.5) grasp the adaptability of variance analysis in practical application 1. one-Factor Analysis of Variance: in the single-factor analysis of variance, if the number of repeated observations in each group is the same or the population is normally distributed, the variance analysis model has a certain bearing capacity, the result is stable only when the ratio of the maximum variance to the minimum variance is less than 3. 2. no repeated multi-factor variance analysis in cells: normality and variance homogeneity are not considered, because normality and variance homogeneity are the basic unit of cells, and each cell has only one data, therefore, it cannot be analyzed. 3. Multi-factor variance analysis with duplicate data in cells: Generally, the data volume is small, so Normality Test and variance uniformity test have no practical significance. (5) simple correlation analysis 1. the Pearson method requires that all variables follow the normal distribution. the non-parametric method is suitable for variables that do not obey the normal distribution. PS: partial correlation analysis and semi-correlation analysis both require normal distribution (Pearson method) (6) prerequisites for Linear Regression Analysis 1. independent Test Methods for independent variables: Multiple linearity test, test indicators are tolerance and variance expansion factor (vif) 2. the residual data is independent and subject to the normal distribution test method. The first method is the plotting method, the second is the dw (Durbin-Watson) test, and the third is the runs test. the relationship between the independent variables and the dependent variables is a linear test method. The first is the plot method, the second is the t-test, and the third is the F test and the final coefficient (7) the use of various t-test. one sample T-test (one sample T test) checks whether the average value of a group of samples is equal to a certain value. independent Samples T-test (independent-sample T test) samples x1, x2 ,..., XN and Sample Y1, Y2 ,..., yn can reverse the order without affecting the result. paired sample T-test sample x1, x2 ,..., XN and sample Y1, Y2 ,..., the order of YN cannot be reversed.