linear regression

linear regression一般指线性回归。 线性回归是利用数理统计中回归分析，来确定两种或两种以上变量间相互依赖的定量关系的一种统计分析方法，运用十分广泛。其表达形式为y = w'x+e，e为误差服从均值为0的正态分布。回归分析中，只包括一个自变量和一个...

Linear Regression

• Sample variance
• Sample covariance
• Sample correlation
• Linear regression
• Multiple regression
• Level of significance
• Critical value
• Model evaluation
  • SEE
  • R-squared
  • ANOVA & F-Test
• Applications and Issues in Regressions
  • Confidence Interval of a Predicted Value
  • Indicator Variables
  • Qualitative Dependent Variables
    • Logit regression
    • Probit regression
    • Discriminant Analysis
  • Limitations of Regression Analysis
  • Assumptions of Regression Models
  • Violation(线性回归模型存在异方差性）
    • Heteroskedasticity
      • Unconditionalheteroskedasticity
      • Conditional heteroskedasticity
      • Correction of Heteroskedasticity
        • robust (heteroskedasticity-consistent) standard errors
        • Generalized Least Squares (GLS)
        • Most statistical software packages
    • Serial Correlation
      • Positive Serial Correlation
      • Negative Serial correlation
      • Test for Serial Correlation
        • Durbin–Watson test
      • Correction for serial correlation
        • Most standard statistical packages
        • Newey-West (1987) method
        • Hansen’s (1982) method
        • Newey&hansen can also use in correcting conditional heteroskedasticity in errors
    • Multicollinearity
      • Correction
        • drop the correlated variables and keep only one
  • Model Misspecifications
    • Omitted-Variable Bias
      • Use log(.) to mitigate the influence of extreme values (outliers) while preserving the order of magnitudes
    • Variables in misspecified functional forms
    • Pooling data
    • Measurement Error Problem
    • Failure to account for autocorrelation in regression errors
      • Time Series Analysis
• Time-Series Data Analysis
  • Trend Analysis
    • Linear trend
    • Log-linear trend
    • Limitation of Trend Models
  • AutoRegressive Time-Series Models
  • 子主题