sbconf(scaleboot)
sbconf()所属R语言包：scaleboot
Bootstrap Confidence Intervals
Bootstrap置信区间
译者：生物统计家园网 机器人LoveR
描述----------Description----------
A confidence interval for a scalar parameter is obtained by inverting the approximately unbiased p-value. This function is very slow, and it is currently experimental.
一个标量参数的置信区间是通过反相约公正的p值。这个功能是非常缓慢的，它是目前处于试验阶段。
用法----------Usage----------
sbconf(x, ...)
## Default S3 method:[默认方法]
sbconf(x,sa, probs=c(0.05,0.95), model="poly.2",
k=2,s=1,sp=-1, cluster=NULL,...)
## S3 method for class 'sbconf':
sbconf(x, probs=x$probs,model=x$model,
k=x$k,s=x$s,sp=x$sp, nofit=FALSE, ...)
## S3 method for class 'sbconf':
plot(x,model=x$model,k=x$k,s=x$s,sp=x$sp,
models = attr(x$fits,"models"), log.xy = "",
xlab="test statistic",ylab=NULL, type.plot = c("p","l","b"),
yval=c("aic","zvalue","pvalue"), sd=2,add=FALSE, col=1:6,
pch=NULL,lty=1:5,lwd=par("lwd"), mk.col=col[1],
mk.lwd=lwd[1], mk.lty=lty[1], ...)
参数----------Arguments----------
参数：x
an object used to select a method. For sbconf.default, x is a list vector of size length{sa} with each element being a vector of bootstrap replicates of a statistic or a list vector of a scalar component.
使用的对象选择方法。对于sbconf.default，x是一个矢量的大小length{sa}的每个元素是一个向量，引导复制的统计或列表的标量向量。
参数：...
further arguments passed to or from other methods.
进一步的参数传递给其他方法。
参数：sa
vector of scales in sigma squared (σ^2).
向量的尺度sigma平方（σ^2“）。
参数：probs
a vector of probabilities at which p-values are inverted.
p-值被反转的概率的向量。
参数：model
a character to specify a model for an AU p-value. This should be included in sboptions("models"), for which model fitting is made internally.
一个字符指定一个模型的AU p值。这应包括在sboptions("models")，内部模型拟合。
参数：k
a numeric to specify an order of AU p-value.
一个数字，以指定的顺序AU p值。
参数：s
σ_0^2
σ_0^2
参数：sp
σ_p^2
σ_p^2
参数：cluster
snow cluster object which may be generated by function makeCluster.
snow的聚类对象可能产生的功能makeCluster。
参数：nofit
logical. No further calls to sbfit are made.
逻辑。没有进一步的调用sbfit。
参数：models
AIC values are plotted for these models.
这些模型的AIC值绘制。
参数：log.xy
character string to be passed to plot.default.
个字符的字符串传递给plot.default。
参数：xlab
a label for the x axis.
为x轴的标签。
参数：ylab
a label for the y axis.
为y轴的标签。
参数：type.plot
a character to be passed to plot.default.
要传递给plot.default字符。
参数：yval
determines y-axis. "aic" for AIC values of models, "zvalue" for AU corrected z-values, and "pvalue" for AU corrected p-values.
确定y轴。 AIC值的模型，“Z值为”AU“AIC”纠正z值，“P值”AU校正P值。
参数：sd
If positive, draws curves +-sd*standard error for z-values and p-values.
如果是正数，绘制曲线+-SD * z值和p值的标准误差。
参数：add
logical. Should not the frame be drawn?
逻辑。如果不是框架划分？
参数：col
vector of colors of plots.
向量的颜色图。
参数：pch
vector of pch's of plots.
向量的PCH的图。
参数：lty
vector of lty's of plots.
LTY向量的图。
参数：lwd
numeric of lwd of plots.
图随钻测井数字。
参数：mk.col
color for crosses drawn at probs.
颜色绘制probs的叉乘。
参数：mk.lwd
lwd for crosses drawn at probs.
LWD叉乘绘制probs。
参数：mk.lty
lty for crosses drawn at probs.
LTY叉乘绘制probs。
Details
详细信息----------Details----------
Let x[[i]] be a vector of bootstrap replicates for a statistic with scale sa[i]. For a threshold value y, the bootstrap probability is bp[i]=sum(x[[i]]<y)/length(x[[i]]). sbconf computes bp for several y values, and finds a value y at which the AU p-value, given by sbfit, equals a probability value specified in probs. In this manner, AU p-values are inverted to obtain bootstrap confidence intervals.
让我们x[[i]]是一个向量，引导规模sa[i]重复统计。阈值y，引导概率是bp[i]=sum(x[[i]]<y)/length(x[[i]])。 sbconf计算bp几个y值，和发现的值y在非盟P-值给出的sbfit，等于指定的概率值在probs。以这种方式，AU p-值被反转，得到自举置信区间。
See the examples below for details.
的详细信息，请参阅下面的例子。
值----------Value----------
sbconf method returns an object of class "sbconf".
sbconf方法返回一个对象类"sbconf"。
The print method for an object of class "sbconf" prints the confidence intervals.
print一个对象的类"sbconf"的方法打印的置信区间。
（作者）----------Author(s)----------
Hidetoshi Shimodaira
参见----------See Also----------
scaleboot.
scaleboot。
实例----------Examples----------
## Not run: [＃不运行：]
## An example to calculate confidence intervals[＃一个例子来计算置信区间]
## The test statistic is that for "t4" in data(mam15)[＃检验统计量是“T4”（mam15）]
## In the following, we used 40 cpu's.[＃在下面，我们用40 CPU。]
##[＃]
library(snow)
library(scaleboot)
cl <- makeCluster(40)
data(mam15)
## Definition of a test statistic of interest.[＃定义的检验统计量的兴趣。]
## "myfun" returns the maximum difference of log-likelihood value[“myfun”返回的最大差值对数似然值]
## for a tree named a.[＃为树命名的。]
myfun <- function(x,w,a) maxdif(wsumrow(x,w))[[a]]
maxdif <- function(x) {
i1 <- which.max(x) # the largest element[最大的元素]
x <- -x + x[i1]
x[i1] <- -min(x[-i1]) # the second largest value[第二大值]
x
}
wsumrow <- function(x,w) {
apply(w*x,2,sum)*nrow(x)/sum(w)
}
clusterExport(cl,c("maxdif","wsumrow"))
## multiscale bootstrap parameters[＃多尺度引导参数]
nb <- 10000
sa <- 10^seq(-2,2,length=13)
## Compute multiscale bootstrap replicates[＃计算多尺度引导复制]
## (It took 80 secs using 40 cpu's)[＃（我花了80秒40 CPU的）]
sim <- scaleboot(mam15.mt,nb,sa,myfun,"t4",count=FALSE,
cluster=cl,onlyboot=TRUE,
names.hp=na,nofit=nofit,models=models)
## Modify option "probs0" to a fine grid with 400 points[＃修改选项“probs0”的细网格，与400点]
## default: 0.001 0.010 0.100 0.900 0.990 0.999[＃默认值：0.001 0.010 0.100 0.900 0.990 0.999]
## NOTE: This modification is useful only when cl != NULL,[注：此修改是很有用的，只有当CL = NULL，]
## in which case calls to sbfit for the grid points[＃在这种情况下，调用以sbfit网格点]
## are made in parallel, although iterations seen later[＃平行，虽然购买看到的迭代]
## are made sequentially.[＃顺序。]
sboptions("probs0",pnorm(seq(qnorm(0.001),qnorm(0.999),length=400)))
## Calculate bootstrap confidence intervals using "k.1" p-value.[＃计算引导使用“K.1”p值的置信区间。]
## (It took 70 secs using 40 cpu's)[（花了70秒，使用40个CPU的）]
## First, sbfit is applied to bp's determined by option "probs0"[＃首先，sbfit应用于到BP的确定选项“probs0”]
## Then, additional fitting is made only twice for iteration.[＃然后，额外的配件只有两次迭代。]
## p[1]=0.05 iter=1 t=4.342723 e=0.0003473446 r=0.0301812[＃P [1] = 0.05 ITER = 1 T = 4.342723 = 0.0003473446 R = 0.0301812]
## p[2]=0.95 iter=1 t=42.76558 e=0.002572495 r=0.1896809[＃p [2] = 0.95 ITER = 1 T = 42.76558 E = 0.002572495 R = 0.1896809]
conf1 <- sbconf(sim$stat,sim$sa,model="sing.3",k=1,cluster=cl)
## The confidence interval with "k.1" is printed as[＃“K.1”打印的置信区间]
## 0.05 0.95 [＃0.05 0.95]
## 4.342723 42.765582 [＃4.342723 42.765582]
conf1
## Calculate bootstrap confidence intervals[＃计算自举置信区间]
## using "k.2" and "k.3" p-values.[＃使用“K.2”和“K.3”p-值。]
## (It took only 10 secs)[＃（只用了10秒）]
## p[1]=0.05 iter=1 t=-2.974480 e=0.003729190 r=0.04725755[＃P [1] = 0.05 ITER = 1 T = -2.974480 E = 0.003729190 R = 0.04725755]
## p[2]=0.95 iter=1 t=39.51767 e=0.001030929 r=0.06141937[＃p [2] = 0.95 ITER = 1 T = 39.51767 E = 0.001030929 R = 0.06141937]
## 0.05 0.95 [＃0.05 0.95]
## -2.974480 39.517671 [＃-2.974480 39.517671]
conf2 <- sbconf(conf1,model="sing.3",k=2)
conf2
## p[1]=0.05 iter=1 t=-3.810157 e=0.01068678 r=0.08793868[＃P [1] = 0.05 ITER = 1 T = -3.810157 E = 0.01068678 R = 0.08793868]
## p[2]=0.95 iter=1 t=39.32669 e=0.001711107 r=0.09464663[＃p [2] = 0.95 ITER = 1 T = 39.32669 E = 0.001711107 R = 0.09464663]
## 0.05 0.95 [＃0.05 0.95]
## -3.810157 39.326686 [＃-3.810157 39.326686]
conf3 <- sbconf(conf2,model="sing.3",k=3)
conf3
## plot diagnostics[＃图诊断]
plot(conf3) # AIC values for models v.s. test statistic value[模型的AIC值V.S.检验统计量的值]
plot(conf3,yval="zval",type="l") # corrected "k.3" z-value[纠正“K.3”Z-值]
stopCluster(cl)
## End(Not run)[＃（不执行）]
转载请注明:出自 生物统计家园网(http://www.biostatistic.net)。
注：
注1：为了方便大家学习，本文档为生物统计家园网机器人LoveR翻译而成，仅供个人R语言学习参考使用，生物统计家园保留版权。
注2：由于是机器人自动翻译，难免有不准确之处，使用时仔细对照中、英文内容进行反复理解，可以帮助R语言的学习。
注3：如遇到不准确之处，请在本贴的后面进行回帖，我们会逐渐进行修订。
|
发表于 2012-9-29 22:26:25
