重用使用总和的计算度量

Question

我知道一些基本的MDX，但似乎无法创建一个计算度量，该度量使用另一个计算度量的日期集合的总和，该度量也是同一日期集合中的另一个度量。

背景是我试图计算简单线性回归系数的最小二乘估计，截距设置为零。我使用这个公式here来做这件事，这很好。然而，我现在想要计算该系数的标准误差，其涉及估计定义的here的残差平方和。

请参阅我在下面编写的代码。 MDX甚至可以实现这一点吗？

With 

Member [Measures].[Renewed] As
    Sum([Customer Product Status].[Customer Product Status Rollup].&[Renewed], [Measures].[Product Count])

Member [Measures].[Renewed * Product Count] As
    [Measures].[Renewed] * [Measures].[Product Count]

Member [Measures].[Product Count * Product Count] As
    [Measures].[Product Count] * [Measures].[Product Count]

Member [Measures].[Product Count n] As
    Sum([Expiry Date].[Date].Children, 1.0)

Member [Measures].[Product Count Mean] As
    Sum([Expiry Date].[Date].Children, [Measures].[Product Count]) / [Measures].[Product Count n]

Member [Measures].[Renewal Rate] As
    (Sum([Expiry Date].[Date].Children, [Measures].[Renewed * Product Count]))/
    (Sum([Expiry Date].[Date].Children, [Measures].[Product Count * Product Count]))

Member [Measures].[Sum of Squared Residuals] As
    Sum([Expiry Date].[Date].Children, ([Measures].[Renewed] - [Measures].[Renewal Rate] * [Measures].[Product Count]) ^ 2.0)

Member [Measures].[Mean of Squared Residuals] As
    [Measures].[Sum of Squared Residuals] / ([Measures].[Product Count n] - 1)

Member [Measures].[Renewal Rate Standard Error] As
    ([Measures].[Mean of Squared Residuals]/Sum([Expiry Date].[Date].Children, [Measures].[Product Count * Product Count])) ^ 0.5

Member [Measures].[Renewal Rate Upper CI] As
    [Measures].[Renewal Rate] + 2.0 * [Measures].[Renewal Rate Standard Error]

Member [Measures].[Renewal Rate Lower CI] As
    [Measures].[Renewal Rate] - 2.0 * [Measures].[Renewal Rate Standard Error]

Set [Month Name] As 
    Filter(
        [Expiry Date].[Month Name].Children, 
        StrToMember("[Expiry Date].[Month Of Year Name].&[May]", CONSTRAINED) * StrToMember("[Expiry Date].[Year Name].&[2015]", CONSTRAINED)
    )

Set [Historical Months] As
    {[Month Name].Item(0).Lag(25) : [Month Name].Item(0).Lag(1)} * 
    [Expiry Date].[Month Date].Children *
    [Expiry Date].[Month of Year Name].Children *
    [Expiry Date].[Year Name].Children

Select Non Empty {[Measures].[Renewal Rate], 
    [Measures].[Renewal Rate Lower CI],
    [Measures].[Renewal Rate Upper CI]} On 0,
    Non Empty [Historical Months] On 1
From [Product Renewals]

Answer 1

我通过重写[Measures]。[Squared Residuals之和]测量来实现这个目的。如果你扩展方程并从sum函数中取出[Measures]。[Renewal Rate]，那么一切正常。扩展的[Measures]。[Squared Residuals之和]度量现在看起来像这样：

select column1, column2, column3
from mydb
where onecolumn in (number1, number2....number13)

4月的Excel回归结果：

MDX回归结果：