多元统计应用ppt.pptx

;;主成分分析法的基本思想 主成分分析（Principal Component Analysis）是利用降维的思想，将多个变量转化为少数几个综合变量（即主成分），其中每个主成分都是原始变量的线性组合，各主成分之间互不相关，从而这些主成分能够反映始变量的绝大部分信息，且所含的信息互不重叠。; a() As Double方阵 EingenValues() As Double特征根, ‘’‘v() As Double特征向量, ByVal count As Long最大迭代次数, ’‘’ByVal n As Long方阵阶数 Sub CyclicJacobi(a() As Double, EingenValues() As Double, v() As Double, ByVal count As Long, ByVal n As Long) Const tolerance = 0.0000000001 Dim maxcount, selfcount, i, p, q As Long Dim Sumsq, avp(), avq() As Double ReDim avp(n, n), avq(n, n) Dim success As Boolean: success = False maxcount = count selfcount = 0 For p = 1 To n 特征向量初始化 For q = 1 To n v(p, q) = 0 Next v(p, p) = 1 Next; Do selfcount = selfcount + 1 Sumsq = 0 For p = 1 To n 主对角线元素平方和 Sumsq = Sumsq + a(p, p) * a(p, p) Next MsgBox 主对角线元素平方和= Sumsq success = True For p = 1 To n - 1 中心，JACOBI计算 For q = p + 1 To n If Abs(a(p, q) / Sumsq) tolerance Then success = False jacobirotation a, p, q, v, n End If Next Next Loop While Not (success Or selfcount maxcount) If success Then For p = 1 To n EingenValues(p) = a(p, p) Next End If End Sub;刘艳博士耐盐试验数据分析;根据以下特征向量与下页计算程序计算主成分得分;Function countCompoScore(v() As Double, x() As Double, ByVal nx As Integer, _ ByVal NSAMPLES As Integer, ByVal nv As Integer, score() As Double, _ ave() As Double, var() As Double, ByVal isCorrelation As Boolean) As Integer Dim i As Integer, j As Integer, k As Integer If isCorrelation Then 相关阵分析结果 For i = 1 To NSAMPLES For j = 1 To nv score(i, j) = 0 For k = 1 To nx score(i, j) = score(i, j) + (x(i, k) - ave(k)) / var(k) * v(k, j)