_覓 | 覦覈襦 | 豕蠏手 | 殊螳 | 譯殊碁 |
FrontPage › 覲螳蟯覿
|
|
[edit]
1 蟯覿(Correlation Analysis) #
[edit]
2 一危 #襴(girth) 覈碁願(weigth)
tmp <- textConnection( "girth weight 35.00 3.45 32.00 3.20 30.00 3.00 31.50 3.20 32.70 3.30 30.00 3.20 36.00 3.85 30.50 3.15 34.70 3.65 30.50 3.40 33.00 3.50 35.00 4.00 31.80 3.10 38.00 4.20 33.00 3.45") x <- read.table(tmp, header=TRUE) close.connection(tmp) plot(x) 一
[edit]
3 螻給(Covariance) #> cov(x) girth weight girth 5.7183810 0.7461667 weight 0.7461667 0.1210238螻給一 企襦 蟯蟯螻襯 螳讌 [edit]
4 殊伎 蟯螻 #> cor.test(girth, weight, data=x, method="pearson") #default: pearson Pearson's product-moment correlation data: girth and weight t = 7.3142, df = 13, p-value = 5.881e-06 alternative hypothesis: true correlation is not equal to 0 95 percent confidence interval: 0.7116680 0.9655589 sample estimates: cor 0.896941
[edit]
5 ろ殊企 蟯螻 #> cor.test(girth, weight, data=x, method = "spearman") Spearman's rank correlation rho data: girth and weight S = 70.0628, p-value = 1.963e-05 alternative hypothesis: true rho is not equal to 0 sample estimates: rho 0.8748878 Warning message: In cor.test.default(girth, weight, data = x, method = "spearman") : Cannot compute exact p-values with ties
[edit]
6 貅 蟯螻 #> cor.test(girth, weight, data=x, method = "kendall") Kendall's rank correlation tau data: girth and weight z = 3.7508, p-value = 0.0001762 alternative hypothesis: true tau is not equal to 0 sample estimates: tau 0.7425743 Warning message: In cor.test.default(girth, weight, data = x, method = "kendall") : ties 覓語 p螳 螻壱 螳 給
[edit]
7 data frame 蟯螻 る蠍 #install.packages("Hmisc") library("Hmisc") rcorr(as.matrix(x), type="pearson") rcorr(as.matrix(x), type="spearman") [edit]
9 螳2 #豢豌: R Graphics Cookbook
panel.cor <- function(x, y, digits=2, prefix="", cex.cor, ...) { usr <- par("usr") on.exit(par(usr)) par(usr = c(0, 1, 0, 1)) r <- abs(cor(x, y, use="complete.obs")) txt <- format(c(r, 0.123456789), digits=digits)[1] txt <- paste(prefix, txt, sep="") if(missing(cex.cor)) cex.cor <- 0.8/strwidth(txt) text(0.5, 0.5, txt, cex = cex.cor * (1 + r) / 2) } panel.hist <- function(x, ...) { usr <- par("usr") on.exit(par(usr)) par(usr = c(usr[1:2], 0, 1.5) ) opar=par(ps=30) #font-size h <- hist(x, plot = FALSE) breaks <- h$breaks nB <- length(breaks) y <- h$counts y <- y/max(y) rect(breaks[-nB], 0, breaks[-1], y, col="white", ...) } panel.lm <- function (x, y, col = par("col"), bg = NA, pch = par("pch"), cex = 1, col.smooth = "black", ...) { points(x, y, pch = pch, col = col, bg = bg, cex = cex) abline(stats::lm(y ~ x), col = col.smooth, ...) #襦覯ろ蟇 #abline(MASS::rlm(y ~ x), col = col.smooth, ...) } pairs(c2009[,2:5], pch=".", upper.panel = panel.cor, diag.panel = panel.hist, lower.panel = panel.lm) [edit]
10 sql襦 蟯螻 蟲蠍 #select (count(*) * sum(x * y) - sum(x) * sum(y)) / (sqrt(count(*) * sum(x * x) - sum(x) * sum(x)) * sqrt(count(*) * sum(y * y) - sum(y) * sum(y))) r from (select x,y from data) t [edit]
11 襦覯ろ 蟯 覿 #library("robust") covRob(stackloss) #Robust Estimate of Covariance covRob(stackloss, corr = TRUE) #Robust Estimate of Correlation 蟆郁骸
> covRob(stackloss, estim = "mcd") Call: covRob(data = stackloss, estim = "mcd") Robust Estimate of Covariance: Air.Flow Water.Temp Acid.Conc. stack.loss Air.Flow 77.32 24.02 54.91 67.30 Water.Temp 24.02 18.28 18.40 25.16 Acid.Conc. 54.91 18.40 99.03 48.99 stack.loss 67.30 25.16 48.99 60.93 Robust Estimate of Location: Air.Flow Water.Temp Acid.Conc. stack.loss 56.15 20.23 85.38 13.15 > covRob(stackloss, corr = TRUE) Call: covRob(data = stackloss, corr = TRUE) Robust Estimate of Correlation: Air.Flow Water.Temp Acid.Conc. stack.loss Air.Flow 1.0000 0.6677 0.6174 0.9514 Water.Temp 0.6677 1.0000 0.4960 0.7868 Acid.Conc. 0.6174 0.4960 1.0000 0.5389 stack.loss 0.9514 0.7868 0.5389 1.0000 Robust Estimate of Location: Air.Flow Water.Temp Acid.Conc. stack.loss 56.92 20.43 86.29 13.73 [edit]
12 ク蟯覿 # 覲螳 蟯蟯螻 レ 譴 3 覲螳 譟伎 朱襦 企ゼ 旧螻 蟯覿
library(ggm) height <- c(160, 155, 170, 160, 180, 177) weight <- c(55, 50, 56, 70,80,73) gender <- c(0,0,0,1,1,1) mydata <- data.frame(height, weight, gender) pcor(c("height", "weight", "gender"), var(mydata)) # partial corr between [height] and [weight] controlling for [gender] cor(height, weight) 蟆郁骸
> pcor(c("height", "weight", "gender"), var(mydata)) [1] 0.8267963 > cor(height, weight) [1] 0.7598541 > --http://wolfpack.hnu.ac.kr/2015_Fall/MDA/MDA_correlation.pdf
ds=read.csv("SMSA_USA.csv") #一危 所鍵 attach(ds) #一危 ; names(ds) #覲企 襴ろ library(car) # 一 れ れ scatterplot.matrix(~d_index+edu+non_white+wc_ratio+income) library(Hmisc) #蟯螻 螻 れ れ rcorr(as.matrix(ds[6:13]),type=spearman") #覿覿蟆一螻 summary(lm(d_index~income)) #襷襯(Y)=(X) 蠏覿 fit=lm(d_index~edu) # 蠏覿 Y=Z(旧覲, 蟲′譴) fit2=lm(income~edu) # 蠏覿 X=Z plot(fit$residuals, fit2$residuals) #一 (X, Y) : Z 旧 summary(lm(fit$residuals~fit2$residuals)) #旧 Y=X 蠏覿 cor.test(fit$residuals,fit2$residuals) #覿覿蟯螻
鏤
|
語 襦 襦 蟆 襴 襦螳 . 襦 蠍語 螻 るГ蠍磯 螻 襦 襷るジ 蠍語 譬蠍磯 覩碁 螳. (螳 豺讀襯 蟆殊蟾? 譴) |