diff --git a/assets/1604925566590.png b/assets/1604925566590.png new file mode 100644 index 0000000..ee5e8db Binary files /dev/null and b/assets/1604925566590.png differ diff --git a/assets/1604925978609.png b/assets/1604925978609.png new file mode 100644 index 0000000..bc41ff1 Binary files /dev/null and b/assets/1604925978609.png differ diff --git a/assets/1604926073407.png b/assets/1604926073407.png new file mode 100644 index 0000000..5d9dd89 Binary files /dev/null and b/assets/1604926073407.png differ diff --git a/assets/1604926082102.png b/assets/1604926082102.png new file mode 100644 index 0000000..05113b4 Binary files /dev/null and b/assets/1604926082102.png differ diff --git a/assets/1604926406229.png b/assets/1604926406229.png new file mode 100644 index 0000000..e83e700 Binary files /dev/null and b/assets/1604926406229.png differ diff --git a/assets/1604926439380.png b/assets/1604926439380.png new file mode 100644 index 0000000..af7d4e1 Binary files /dev/null and b/assets/1604926439380.png differ diff --git a/assets/1604926455663.png b/assets/1604926455663.png new file mode 100644 index 0000000..29cfad3 Binary files /dev/null and b/assets/1604926455663.png differ diff --git a/assets/1604926463864.png b/assets/1604926463864.png new file mode 100644 index 0000000..8037a5a Binary files /dev/null and b/assets/1604926463864.png differ diff --git a/assets/1604926496771.png b/assets/1604926496771.png new file mode 100644 index 0000000..8037a5a Binary files /dev/null and b/assets/1604926496771.png differ diff --git a/assets/1604926508213.png b/assets/1604926508213.png new file mode 100644 index 0000000..5ca433e Binary files /dev/null and b/assets/1604926508213.png differ diff --git a/assets/1604926618861.png b/assets/1604926618861.png new file mode 100644 index 0000000..29cc2fd Binary files /dev/null and b/assets/1604926618861.png differ diff --git a/assets/1604926821269.png b/assets/1604926821269.png new file mode 100644 index 0000000..0c85d79 Binary files /dev/null and b/assets/1604926821269.png differ diff --git a/必备数学基础.md b/必备数学基础.md index 4730356..db45f22 100644 --- a/必备数学基础.md +++ b/必备数学基础.md @@ -1037,7 +1037,7 @@ H0:β1 = 0所有回归系数与零无显著差异,y与全体x的线性关系 - 在一元线性回归中,自变量只有一个,线性关系检验与回归系数检验是等价的 - 多元回归分析中,这两种检验的意义是不同的。线性关系检验只能用来检验总体回归关系的显著性,而回归系数检验可以对各个回归系数分别进行检验 -**多元线性回归分析** +#### 多元与曲线回归问题 经常会遇到某一现象的发展和变化取决于几个影响因素的情况,也就是一个因变量和几个自变量有依存关系的情况,这时需用多元线性回归分析。 @@ -1047,3 +1047,87 @@ H0:β1 = 0所有回归系数与零无显著差异,y与全体x的线性关系 ![1604846770738](assets/1604846770738.png) +**调整的多重判定系数**: + +用样本容量n和自变量的个数k去修正R^2得到: + +![1604925566590](assets/1604925566590.png) + +- 避免增加自变量而高估R^2 + +**曲线回归分析**: + +直线关系是两变量间最简单的一种关系,曲线回归分析的基本任务是通过两个相关变量x与y的实际观测数据建立曲线回归方程,以揭示x与y间的曲线联系的形式。 + +曲线回归分析最困难和首要的工作是确定自变量与因变量间的曲线关系的类型,曲线回归分析的基本过程: + +- 先将x或y进行变量转换 +- 对新变量进行直线回归分析、建立直线回归方程并进行显著性检验和区间估计 +- 将新变量还原为原变量,由新变量的直线回归方程和置信区间得出原变量的曲线回归方程和置信区 + +由于曲线回归模型种类繁多,所以没有通用的回归方程可直接使用。但是对于某些特殊的回归模型,可以通过变量代换、取对数等方法将其线性化,然后使用标准方程求解参数,再将参数带回原方程就是所求。 + +**实例**:**某商店各个时期的商品流通费率和商品零售额资料** + +| 商品零售额x(万元) | 9.5 | 11.5 | 13.5 | 15.5 | 17.5 | 19.5 | 21.5 | 23.5 | 25.5 | 27.5 | +| ------------------- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | +| 商品流通费率y(%) | 6 | 4.6 | 4 | 3.2 | 2.8 | 2.5 | 2.4 | 2.3 | 2.2 | 2.1 | + +散点图如下: + +![1604925978609](assets/1604925978609.png) + +散点图显示出x与y的变动关系为一条递减的双曲线。 + +![1604926073407](assets/1604926073407.png) + +![1604926082102](assets/1604926082102.png) + +> 这样转换后,公式和线性公式是一样的 + +标准方程为![1604926406229](assets/1604926406229.png) + +将计算数据代入![1604926439380](assets/1604926439380.png) + +解得![1604926455663](assets/1604926455663.png) + +![1604926463864](assets/1604926463864.png)= -0.4377 + 60.4x' + +x' = 1/x 代入 + +![1604926496771](assets/1604926496771.png)=-0.4377+![1604926508213](assets/1604926508213.png) + +**多重共线性** + +回归模型中两个或两个以上的自变量彼此相关的现象 + +举例 x1={10,8,6,4},x2={5,4,3,2},那么x1是两倍的x2 + + + +多重共线性带来的问题有: + +- 回归系数估计值的不稳定性增强 +- 回归系数假设检验的结果不显著等 + +**多重共线性检验的主要方法** + +- 容忍度 +- 方差膨胀因子(VIF) + +容忍度: + +![1604926618861](assets/1604926618861.png) + +- Ri是解释变量xi与方程中其它解释变量间的复相关系数; +- 容忍度在0~1之间,越接近于0,表示多重共线性越强,越接近于1,表示多重共线性越弱; + +**方差膨胀因子** + +方差膨胀因子是容忍度的倒数 + +![1604926821269](assets/1604926821269.png) + +- VIFi越大,特别是大于等于10,说明解释变呈xi与方程中其他解释变量之间有严重的多重共线性; +- VIFi越接近1,表明解释变量xi和其他解释变量之间的多重共线性越弱; +