wave.band(waveband)
wave.band()所属R语言包：waveband
Posterior credible intervals for wavelet regression
小波回归后的置信区间为
译者：生物统计家园网 机器人LoveR
描述----------Description----------
Computes posterior credible intervals for an unknown regression curve.
计算后的置信区间为未知的回归曲线。
用法----------Usage----------
wave.band(data = 0, alpha = 0.5, beta = 1., filter.number = 8, family =
"DaubLeAsymm", bc = "periodic", dev = var, j0 = 3., plotfn = TRUE,
retvalue = TRUE, n = 128, type = "data", rsnr = 3)
参数----------Arguments----------
参数：data
If type="data", then data should be a vector of data. The length of the vector should be a power of two not greater than 1024.
如果type =“数据”，然后数据应该是一个矢量数据。该矢量的长度应该是两个不大于1024的电源。
参数：type
Either type="data", in which case a vector of data should be supplied, or type should specify a standard test function and wave.band will generate a test data set via a call to test.data. Permissible values for type are "blocks", "bumps", "doppler", "heavi", or "ppoly"; see the documentation for test.data for more details.
任一类型=“数据”，在这种情况下，应提供的矢量数据，或类型应指定一个标准的测试的功能和wave.band将生成的测试数据通过设置一个呼叫test.data。允许值的类型是“块”，“凸点”，“多普勒”heavi“，或”ppoly“为test.data有关详细信息，请参阅文档。
参数：alpha, beta
Hyperparameters which determine the priors placed on the wavelet coefficients. Both alpha and beta take positive values; see Abramovich, Sapatinas, & Silverman (1998) or Chipman & Wolfson (1999) for more details on selecting alpha and beta.
蔡怀平确定对上的小波系数下的先验概率。 α和β都取正值，选择α和β的详细信息，请参阅阿布拉莫维奇，Sapatinas，和西尔弗曼（1998）或奇普曼欧胜（1999）。
参数：filter.number
A parameter relating to the smoothness of wavelet that you want to use in the decomposition.
有关的参数要使用的分解小波的平滑度。
参数：family
Specifies the family of wavelets to be used. Two popular options are "DaubExPhase" and "DaubLeAsymm" but see the help for filter.select for more possibilities.
指定要使用的小波家庭。两种流行的的选项“DaubExPhase”和“DaubLeAsymm”，但看到更多的可能性filter.select的帮助。
参数：bc
Specifies the boundary handling. If bc="periodic" the default, then the function you decompose is assumed to be periodic on it's interval of definition. Other boundary options exist, but are currently unsupported for this function.
指定的边界处理。如果BC =“定期”的默认，然后分解的功能，您认为是周期性的间隔的定义。其他边界是存在的，但目前不支持此功能。
参数：dev
This argument supplies the function to be used to compute the spread of the absolute values coefficients. The function supplied must return a value of spread on the variance scale (i.e. not standard deviation) such as the var() function. A popular, useful and robust alternative is the madmad function.
此参数提供的功能被用于计算的绝对值的系数的传播。提供的功能必须返回一个值传播的方差规模的（即不标准差），比如VAR（）函数。一个流行的，有用的和强大的的选择是madmad的功能。
参数：j0
The primary resolution level; used in assessing the universal threshold which is used in the empirical Bayes estimation of hyperparameters.
主要的分辨率水平，在评估中所用的经验Bayes估计的超全局阈值。
参数：plotfn
If plotfn=TRUE, wave.band draws the noisy data, the BayesThresh function estimate, and pointwise 99 percent credible intervals for the regression function. If the value of type is not "data", then the true function will also be plotted.
如果plotfn=TRUE，wave.band吸引了喧闹的数据，的BayesThresh函数估计，并逐点99％可信区间为回归函数。 type如果该值是不是“数据”，那么真正的功能也将被绘制。
参数：retvalue
If retvalue=TRUE, then a lengthy list of results will be returned. Note that if both plotfn and retvalue are set to FALSE, then wave.band will return no results whatsoever.
如果retvalue=TRUE，然后是一长串的结果将被返回。需要注意的是，如果没有plotfn和retvalue设置为FALSE，那么wave.band将返回没有任何结果。
参数：n
If type is not "data", then a data vector of length n will be generated; note that n should be a power of two not greater than 1024.
如果type是“数据”，然后数据向量的长度n会产生注意，n应该是功率不大于1024。
参数：rsnr
If type is not "data", then the data vector generated will have root signal-to-noise ratio as specified by rsnr.
如果类型是“数据”，然后将数据矢量根信号与噪声比指定的rsnr。
Details
详细信息----------Details----------
This function implements the WaveBand method of Barber, Nason, & Silverman (2001) to compute posterior credible intervals for a regression function. The credible intervals are found by approximating the posterior distribution of the estimated regression curve at each design point. A mixture prior with two components (a zero-mean normal and a point mass at zero) is placed on each wavelet coefficient and updated by the data to give the posteriors for the wavelet coefficients. This is the same prior used by Abramovich, Sapatinas, & Silverman (1998) in their BayesThresh method, implemented in the function BAYES.THR.
此功能实现的波段理发，利晨和Silverman（2001）的方法计算后的置信区间为回归函数。可信区间近似的后验分布在每一个设计点的估计回归曲线。前的混合物与两种组分（一个零均值的正常点质量为零时）被放置在每个小波系数，并更新由该数据，得到的后验概率的小波系数。这是维奇，Sapatinas和Silverman（1998）在其BayesThresh方法之前所使用的相同，实施中的功能BAYES.THR。
The cumulants of these posteriors are computed and stored in the wd.objects returned by wave.band as Kr.wd. These are summed to give the posterior cumulants of the regression curve, which are used to fit a Johnson distribution (Johnson, 1949), using the algorithm of Hill, Hill, & Holder (1976). Percentage points of these distributions are computed by the algorithm of Hill (1976) and give the credible intervals themselves.
这些后验概率，累积的计算和存储在wd.objects返回wave.bandKr.wd的。这些总结后累积的回归曲线，以适应约翰逊分布（约翰逊，1949年），这是用来使用的算法的希尔，希尔，持有人（1976年）。这些分布计算个百分点山（1976）的算法，并给自己的可信区间。
Code to implement the algorithms by Hill (1976) and Hill, Hill, & Holder (1976) was obtained from the StatLib archive.
山（1976）和希尔，希尔，持有人（1976年）的代码来实现的算法，得到从StatLib存档。
值----------Value----------
If retvalue=FALSE, the value returned by wave.band is NULL. Otherwise, wave.band returns a list with the following components:
如果retvalue=FALSE，返回的值wave.band是NULL。否则，wave.band返回一个列表，以下组件：
<table summary="R valueblock"> <tr valign="top"><td>data</td> <td> The data vector which has been analysed. </td></tr> <tr valign="top"><td>nts</td> <td> A list containing four vectors named one, two, three, and four. Vector one contains the first cumulants of the regression function estimate, vector to the second cumulants and so on.</td></tr> <tr valign="top"><td>Kr.wd</td> <td> A list of four wd objects. These contain the first to fourth cumulants of the wavelet coefficients, as well as recording the wavelet used in the decomposition. </td></tr> <tr valign="top"><td>bands</td> <td> A list containing pointwise upper and lower credible limits for the regression function estimate for nominal coverage rates 80, 90, 95 and 99 percent. The widths of the credible intervals is also included. The vectors are named with "l", "u", and "w" indicating lower limits, upper limits, and intervals widths, while "80", "90", "95", and "99" refer to the nominal coverage rate. </td></tr>
<table summary="R valueblock"> <tr valign="top"> <TD> data</ TD> <TD>的数据向量进行了分析。 </ TD> </ TR> <tr valign="top"> <TD> nts</ TD> <td>一个列表，其中包含四个向量命名为1，2，3，和4。矢量包含的第一个累积的回归函数估计，向量的第二个累积量等。</ TD> </ TR> <tr valign="top"> <TD> Kr.wd </ TD> < TD> A四个WD对象列表。这些包含在第一至第四的小波系数的累积量，以及记录用于分解的小波。 </ TD> </ TR> <tr valign="top"> <TD>bands </ TD> <td>一个列表的回归函数估计的名义覆盖率80包含逐点上下的可信限制，90％，95％和99％。可信区间的宽度也包括在内。该向量被命名为“l”的，“u”的，而“w”，表示下限值，上限值，和间隔的宽度，而“80”，“90”，“95”，和“99”指的标称覆盖率。 </ TD> </ TR>
</table> The BayesThresh estimate of the regression function, using the same parameters as the WaveBand credible intervals, is also included in the pointest component of this list.
</表>BayesThresh回归函数的估计，使用波段的置信区间的相同的参数，也被包括在pointest组件此列表。
<table summary="R valueblock"> <tr valign="top"><td>param</td> <td> A record of parameters in the call to wave.band.</td></tr> </table>
<table summary="R valueblock"> <tr valign="top"> <TD> param</ TD> <td>一个记录中的参数调用wave.band。</ TD> </ TR> </ TABLE>
副作用----------SIDE EFFECTS----------
If plotfn=TRUE, results are plotted on the current graphics device.
如果plotfn=TRUE，结果被绘制在当前的图形设备。
参见----------See Also----------
BAYES.THR, plot.wb, power.sum, test.data
BAYES.THR，plot.wb，power.sum，test.data
实例----------Examples----------
#library(wavethresh)[库（wavethresh）]
#[]
# First, look at the piecewise polynomial example.[首先，在分段多项式的例子。]
#[]
# This plot and the plots for the smooth example below show[这个图和图的顺利下面的例子展示]
# the data as points, the BayesThresh estimate (thick line), [点的数据作为，BayesThresh估计（粗线），]
# pointwise 99 percent credible intervals (thin lines), and[逐点的99％置信区间（细线），和]
# the true function (dotted thin line).[真正的功能（虚线细线）。]
#[]
ppoly.wb <- wave.band(type = "ppoly", n = 1024, rsnr=4)
#[]
# Plotting the cumulants shows that there are significant [绘制的累积量有显着的]
# third and fourth cumulants in some places.[在一些地方的第三个和第四个累积。]
#[]
t <- (1:1024)/1024
plot(t, ppoly.wb$cumulants$one, type="l", xlab="t", ylab = "one")
plot(t, ppoly.wb$cumulants$two, type="l", xlab="t", ylab = "two")
plot(t, ppoly.wb$cumulants$three, type="l", xlab="t", ylab = "three")
plot(t, ppoly.wb$cumulants$four, type="l", xlab="t", ylab = "four")
#[]
# Now consider how much difference the prior can make.[现在，考虑多大的区别之前可以。]
# Consider a smooth example, first using the default prior,[考虑一个平稳的例子，第一次使用之前默认的，]
# and then using a smoother prior.[然后使用之前更顺畅。]
#[]
gs <- sin(2*pi*t) + 2*(t - 0.5)^2
gs.noisy <- gs + rnorm(n=1024, sd=sqrt(var(gs))/2)
gs.wb1 <- wave.band(data=gs.noisy)
gs.wb2 <- wave.band(data=gs.noisy, alpha=4, beta=1)
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 2012-10-1 16:52:06
