Other articles with its own series, put together here
####################### function own, run directly call ################## ### function # of calculating mse mse = function (ei, p) #ei is residual vector, p is the number of the regression parameters { n-= length (EI) SSE = SUM (EI ** 2) mse = SSE / ( NP) return (MSE) } # hat calculated matrix, and extracting the diagonal elements H = function (X) #X regression vector matrix { H *% =% X-Solve (T (X-)%% X-*)% * T% (X-) HII = diag (H) return (HII) } # Kaiman RI function = (EI, X-) { P = ncol (X-) # regression coefficient number mse = mse (ei, p) # residual square hii = H (X) # returns the diagonal elements of the matrix hat ans = ei / sqrt (mse * (1-hii)) # Kaiman return (ANS) } # external Kaiman ti = function (ei, X) # input variable regression residuals matrix { p = ncol (X) # number of regression parameters n = length (ei) # number of data hii = H (X) # Hat main diagonal elements of the matrix s2_i = ((np) * mse (ei, p) - ( ei ** 2) / (1- hii)) / (np-1) # calculates S (I) ^ 2 ANS = EI / sqrt (s2_i * (. 1-HII)) return (ANS) } # statistic calculation PRESS press = function (ei, X) #X is the design matrix argument { HII = H (X-) RES = SUM ((EI / (. 1-HII)) ** 2) #View (RES) } # PRESS is calculated prediction 2 ^ R & lt R_pred = function (X-, Y) { HII = H (X-) EI = RESID (LM (X-Y ~ [, 2] X-+ [,. 3])) PRESS = SUM ((EI / (l- HII)) ** 2) SST = SUM ((Y-Mean (Y)) ** 2) ANS-PRESS. 1 = / SST return (ANS) } # draw a normal probability plot plot_ZP = function (ti) # external input Kaiman { n-length = (Ti) Order = Rank (Ti) # ascending order, t (i) is the first order a Pi = (order-1/2 ) / n # cumulative probability plot (ti, Pi, xlab = " student residuals", ylab = "percent") # Videos normal probability plot # regression line FM LM = (Pi ~ Ti) abline (FM) } # test for lack of fit # Library (rsm) # rsm bag loading test for lack of fit # lm.rsm <-rsm (FO ~ Y (X)) # call loss # loftest (lm.rsm) fit test function loftest # calculation Dii ' Di_i = function (i, I_, MSE, beta1, Beta2, new_data) #i of the i-th point, i_ i_ first point, data dataset { one beta1 = * ($ new_data Cases [I] Cases -new_data $ [I _]) / sqrt (MSE) tWO Beta2 = * (Distance new_data $ [I] - $ Distance new_data [I _]) / sqrt (MSE) ANS = One ** 2 ** 2 + TWO return (ANS) }