EBMTP(multtest)
EBMTP()所属R语言包：multtest
A function to perform empirical Bayes resampling-based multiple hypothesis testing
一个函数来执行经验Bayes重采样为基础的多假设检验
译者：生物统计家园网 机器人LoveR
描述----------Description----------
A user-level function to perform empirical Bayes multiple testing procedures (EBMTP). A variety of t- and F-tests, including robust versions of most tests, are implemented. A common-cutoff method is used to control the chosen type I error rate (FWER, gFWER, TPPFP, or FDR). Bootstrap-based null distributions are available. Additionally, for t-statistics, one may wish to sample from an appropriate multivariate normal distribution with mean zero and correlation matrix derived from the vector influence function. In EBMTP, realizations of local q-values, obtained via density estimation, are used to partition null and observed test statistics into guessed sets of true and false null hypotheses at each round of (re)sampling. In this manner, parameters of any type I error rate which can be expressed as a function the number of false positives and true positives can be estimated. Arguments are provided for user control of output. Gene selection in microarray experiments is one application.
一个用户级的函数来执行经验Bayes多个测试程序（EBMTP）。的各种T-和F-测试，包括强劲大多数测试版本，实施。截止一种常见的方法是用来控制所选择的类型我错误率（或FWER，gFWER，TPPFP，FDR）的。引导为主的空分布。此外，为t-统计，不妨品尝从一个适当的多元正态分布零均值和矢量影响功能的相关矩阵。 EBMTP，当地Q值的实现，获得通过密度估计，用于猜到真假虚无假设套在每一轮（重）抽样分区空和观测试验统计。在这种方式下，任何I型错误率参数，可以作为一个功能表示误报和真阳性的数量可估计。输出控制，为用户提供参数。基因芯片实验选择的是一个应用程序。
用法----------Usage----------
EBMTP(X, W = NULL, Y = NULL, Z = NULL, Z.incl = NULL, Z.test = NULL,
na.rm = TRUE, test = "t.twosamp.unequalvar", robust = FALSE,
standardize = TRUE, alternative = "two.sided", typeone = "fwer",
method = "common.cutoff", k = 0, q = 0.1, alpha = 0.05, smooth.null = FALSE,
nulldist = "boot.cs", B = 1000, psi0 = 0, marg.null = NULL,
marg.par = NULL, ncp = NULL, perm.mat = NULL, ic.quant.trans = FALSE,
MVN.method = "mvrnorm", penalty = 1e-06, prior = "conservative",
bw = "nrd", kernel = "gaussian", seed = NULL, cluster = 1,
type = NULL, dispatch = NULL, keep.nulldist = TRUE, keep.rawdist = FALSE,
keep.falsepos = FALSE, keep.truepos = FALSE, keep.errormat = FALSE,
keep.Hsets=FALSE, keep.margpar = TRUE, keep.index = FALSE, keep.label = FALSE)
参数----------Arguments----------
参数：typeone
Character string indicating which type I error rate to control, by default family-wise error rate ('fwer'). Other options include generalized family-wise error rate ('gfwer'), with parameter k giving the allowed number of false positives, and tail probability of the proportion of false positives ('tppfp'), with parameter q giving the allowed proportion of false positives. The false discovery rate ('fdr') can also be controlled. In particular, for 'gfwer', 'tppfp' and 'fdr', multiple testing is not performed via augmentation of the results of a FWER-controlling MTP. Rather, using guessed sets of true and false null hypotheses, these error rates are controlled in a more direct manner.
字符串指示哪种类型，我的错误率控制，默认家庭明智的错误率（fwer）。其他选项包括家庭明智的广义错误率参数（“gfwer），k给予允许数量的误报，误报（”tppfp）的比例和尾部概率，参数<X >误报允许的比例。也可以控制错误发现率（FDR）。特别是，“gfwer，tppfp”和“FDR”，不执行多个测试通过隆胸的FWER控股的中期计划的结果。相反，使用猜到套真假虚无假设，这些错误率控制在一个更直接的方式。
参数：method
Character string indicating the EBMTP method. Currently only 'common.cutoff' is implemented. This method is most similar to 'ss.maxT' in MTP.
字符串，表示EBMTP方法。目前只有common.cutoff“实施。这种方法是最相似的“ss.maxTMTP。
参数：nulldist
Character string indicating which resampling method to use for estimating the joint test statistics null distribution, by default the non-parametric bootstrap with centering and scaling ('boot.cs'). The old default 'boot' will still compile and will correspond to 'boot.cs'. Other null distribution options include 'boot.ctr', 'boot.qt', and 'ic', corresponding to the centered-only bootstrap distribution, quantile-transformed bootstrap distribution, and influence curve multivariate normal joint null distribution, respectively. The permutation distribution is not available.
字符串表示使用重采样方法估算默认情况下，联合测试空分布统计，与中心和缩放（“boot.cs）非参数引导。旧的默认的引导仍将编译，将符合“boot.cs。其他空分布选项boot.ctr，boot.qt“，”IC“，相应的中心只引导分布，分位数转化的引导分布，曲线多元正常的联合空分布的影响，分别。排列分布是不可用。
参数：prior
Character string indicating which choice of prior probability to use for estimating local q-values (i.e., the posterior probabilities of a null hypothesis being true given the value of its corresponding test statistic). Default is 'conservative', in which case the prior is set to its most conservative value of 1, meaning that all hypotheses are assumed to belong to the set of true null hypotheses. Other options include 'ABH' for the adaptive Benjamini-Hochberg estimator of the number/proportion of true null hypotheses, and 'EBLQV' for the empirical Bayes local q-value value estimator of the number/proportion of true null hypotheses. If 'EBLQV', the estimator of the prior probability is taken to be the sum of the estimated local q-values divided by the number of tests. Relaxing the prior may result in more rejections, albeit at a cost of type I error control under certain conditions. See details and references.
字符串，指示用于估算当地的Q值（即空假说真正给予其相应的检验统计量的值的后验概率）先验概率的选择。默认是“保守”，在这种情况下，事先设置其值为1，最保守的，这意味着所有的假设都假定属于一套真正的零假设。其他选项包括自适应Benjamini Hochberg的估计数/真正的虚无假设比例的“陆地棉”，EBLQV经验Bayes局部Q值/真正的零假设的数量比例值估计。先验概率估计如果“EBLQV，估计数除以测试Q值的总和。放宽之前，可能会导致更多的拒绝，但在I型误差控制在一定条件下的成本。查看详细信息和参考。
参数：bw
A character string argument to density indicating the smoothing bandwidth to be used during kernel density estimation. Default is 'nrd'.
一个字符串参数density表示将在核密度估计的平滑带宽。默认为“NRD”。
参数：kernel
A character string argument to density specifying the smoothing kernel to be used. Default is 'gaussian'.
一个字符串参数density指定要使用平滑内核。默认为“高斯”。
参数：keep.falsepos
A logical indicating whether or not to store the matrix of guessed false positives at each round of (re)sampling. The matrix has rows equal to the number of cut-offs (observed test statistics) and columns equal to the B number of bootstrap samples or samples from the multivariate normal distribution (if nulldist='ic'). Default is 'FALSE'.
一个逻辑说明是否存储在每一轮（重）抽样矩阵猜到误报。矩阵行等于截止权衡（观察测试统计）和列等于B引导多元正态分布的样品或样品的数量，（如果nulldist='ic'）。默认为“假”。
参数：keep.truepos
A logical indicating whether or not to store the matrix of guessed true positives at each round of (re)sampling. The matrix has rows equal to the number of cut-offs (observed test statistics) and columns equal to the B number of bootstrap samples or samples from the multivariate normal distribution (if nulldist='ic'). Default is 'FALSE'.
一个逻辑说明是否存储在每一轮（重）抽样猜到真阳性的矩阵。矩阵行等于截止权衡（观察测试统计）和列等于B引导多元正态分布的样品或样品的数量，（如果nulldist='ic'）。默认为“假”。
参数：keep.errormat
A logical indicating whether or not to store the matrix of type I error rate values at each round of (re)sampling. The matrix has rows equal to the number of cut-offs (observed test statistics) and columns equal to the B number of bootstrap samples or samples from the multivariate normal distribution (if nulldist='ic'). Default is 'FALSE'. In the case of FDR-control, for example, this matrix is falsepos/(falsepos + truepos). The row means of this matrix are eventually used for assigning/ordering adjusted p-values to test statistics of each hypothesis.
一个逻辑说明是否存储I类错误率值的矩阵，在每一轮（重）抽样。矩阵行等于截止权衡（观察测试统计）和列等于B引导多元正态分布的样品或样品的数量，（如果nulldist='ic'）。默认为“假”。例如，在FDR控制的情况下，这个矩阵是falsepos/（falsepos+truepos）。该行指最终用于分配/排序调整p值来测试每一个假设的统计，这个矩阵。
参数：keep.Hsets
A logical indicating whether or not to return the matrix of indicators which partition the hypotheses into guessed sets of true and false null hypotheses at each round of (re)sampling. Default is 'FALSE'.
一个逻辑指示是否返回矩阵的指标将在每一轮（重）抽样真假虚无假设猜测套分区的假设。默认为“假”。
参数：X, W, Y, Z, Z.incl, Z.test, na.rm, test, robust, standardize, alternative, k, q, alpha, smooth.null, B, psi0, marg.null, marg.par, ncp, perm.mat, ic.quant.trans, MVN.method, penalty, seed, cluster, type, dispatch, keep.nulldist, keep.rawdist, keep.margpar, keep.index, keep.label
These arguments are all similarly used by the MTP function, and their use has been defined elsewhere. Please consult the link{MTP} help file or the references for further details. Note that the MTP-function arguments get.cr, get.cutoff, get.adjp are now DEPRECATED in the EBMTP function. Only adjusted p-values are calculated by EBMTP. These adjusted p-values are returned in the same order as the original hypotheses and raw p-values (typically corresponding to rows of X.)
这些论点都同样使用MTP功能，其使用已在别处定义。请咨询link{MTP}帮助文件或进一步的细节参考。请注意MTP-函数的参数get.cr, get.cutoff, get.adjpEBMTP函数现在已过时。只调整p值计算EBMTP。这些调整后的P-值返回原假设和原材料p值（通常对应行X。）以相同的顺序
Details
详情----------Details----------
The EBMTP begins with a marginal nonparametric mixture model for estimating local q-values. By definition, q-values are 'the opposite' of traditional p-values. That is, q-values represent the probability of null hypothesis being true given the value of its corresponding test statistic. If the test statistics Tn have marginal distribution f = pi*f_0 + (1-pi)f_1, where pi is the prior probability of a true null hypothesis and f_0 and f_1 represent the marginal null and alternative densities, respectively, then the local q-value function is given by pi*f_0(Tn)/f(Tn). <br>
EBMTP开始与边际非参数混合模型估计当地Q-值。根据定义，Q值是传统的P-值“对面”。也就是说，Q值代表空是真实的假设给予其相应的检验统计量的值的概率。如果测试统计TN边际分布F = PI * F_0 +（1-π）F_1，其中pi是一个真正的零假设的先验概率和F_0和F_1代表边际空和替代密度，分别，然后局部q值函数PI * F_0（TN）/ F（TN）。参考
One can estimate both the null density f_0 and full density f by applying kernel density estimation over the matrix of null test statistics and the vector of observed test statistics, respectively. Practically, this step in EBMTP also ensures that sidedness is correctly accounted for among the test statistics and their estimated null distribution. The prior probability pi can be set to its most conservative value of 1 or estimated by some other means, e.g., using the adaptive Benjamini Hochberg ('ABH') estimator or by summing up the estimated local q-values themselves ('EBLQV')and dividing by the number of tests. Bounding these estimated probabilities by one provides a vector of estimated local q-values with length equal to the number of hypotheses. Bernoulli 0/1 realizations of the posterior probabilities indicate which hypotheses are guessed as belonging to the true set of null hypotheses given the value of their test statistics. Once this partitioning has been achieved, one can count the numbers of guessed false positives and guessed true positives at each round of (re)sampling that are obtained when using the value of an observed test statistic as a cut-off. <br>
可以估计申请空试验的统计和观察测试统计向量矩阵核密度估计空密度F_0和全密度f，分别。实际上，这一步EBMTP也确保片面性正确占其中的检验统计量和他们的估计空分布。先验概率PI可以设置其最保守的1值或估计其他一些手段，如使用的，自适应Benjamini Hochberg（“陆地棉”）估计，或通过总结估计当地Q-值本身（EBLQV）除以测试。这些估计的概率边界长度等于假设的数量估计当地Q值提供了一个向量。伯努利0/1的后验概率的实现表明，假设属于其检验统计量的值的零假设的真实猜到。一旦这个分区已经实现了，我们可以计算的数字猜测误报和猜测在每一轮（重）采样得到使用时观察到的检验统计量的值作为截止真阳性。参考
EBMTPs use function closures to represent type I error rates in terms of their defining features. Restricting the choice of type I error rate to 'fwer', 'gfwer', 'tppfp', and 'fdr', means that these features include whether to control the number of false positives or the proportion of false positives among the number of rejetions made (i.e., the false discovery proportion), whether we are controlling a tail probability or expected value error rate, and, in the case of tail probability error rates, what bound we are placing on the random variable defining the type I error rate (e.g., k for 'gfwer' or 'q' for 'tppfp'). Averaging the type I error results over B (bootstrap or multivariate normal) samples provides an estimator of the evidence against the null hypothesis (adjusted p-values) with respect to the choice of type I error rate. Finally, a monotonicity constraint is placed on the adjusted p-values before being returned as output. <br>
EBMTPs使用功能关闭，我在其定义的功能方面的错误率代表类型。限制选择I型错误率fwer，gfwer，tppfp“，”FDR“，意味着这些功能包括是否控制误报数量或误报的比例之间的rejetions （即假的发现比例），我们是否有尾巴概率或预期值的误差率控制，并在尾概率的错误率的情况下，我们什么必然放置随机变量定义的I型错误率（例如，gfwer或qtppfp）K。平均比B I型错误的结果（引导或多元正常）样本提供的证据对零假设的选择I型错误率（p-值调整）的估计。最后，单调性约束被放置在调整后的P-值之前被作为输出返回。参考
As detailed in the references, relaxing the prior may result in a more powerful multiple testing procedure, albeit sometimes at the cost of type I error control. Additionally, when the proportion of true null hypotheses is close to one, type I error control may also become an issue, even when using the most conservative prior probability of one. This feature is known to occur with some other procedures which rely on the marginal nonparametric mixture model for estimating (local) q-values. The slot EB.h0M returned by objects of class EBMTP is the sum of the local q-values estimated via kernel density estimation (divided by the total number of tests). If this value is close to one (>0.9-0.95), the user will probably not want relax the prior, as even the conservative EBMTP might be approaching a performance bound with respect to type I error control. The user is advised to begin by using the most 'conservative' prior, assess the estimated proportion of true null hypotheses, and then decide if relaxing the prior might be desired. Gains in power over other multiple testing procedures have been observed even when using the most conservative prior of one. <br>
至于详细的引用，放宽之前，可能会导致在一个更强大的多个测试程序，虽然有时，在I型误差控制成本。此外，真正的零假设的比例接近1时，输入错误的控制也可能成为一个问题，即使使用最保守的一个先验概率。这一功能被称为会出现一些边际非参数混合模型估计Q值（本地）依靠其他程序。插槽EB.h0M类对象返回EBMTP是当地通过核密度估计除以测试总数估计Q值的总和。如果这个值接近（0.9-0.95），用户可能会不希望放松之前，因为即使是的保守EBMTP可能接近键入我的错误控制方面的约束性能。建议用户使用最“保守”的前开始，真正的零假设评估估计比例，然后决定如果放宽之前，可能会所需。涨幅超过其他多个测试程序的权力已经观察到，即使在使用前最保守的一个。参考
Situations of moderate-high to high levels of correlation may also affect the results of multiple testing methods which use the same mixture model for generating q-values. Microarray analysis represents a scenario in which dependence structures are typically weak enough to mitigate this concern. On the other hand, the analysis of densely sampled SNPs, for example, may present problems. <br>
中等高高层次的相关情况也可能会影响多个使用相同的混合模型生成Q值的测试方法的结果。基因芯片分析代表一个场景中的依赖结构通常是弱，不足以减轻这种担忧。另一方面，密集采样的单核苷酸多态性分析，例如，可能会出现问题。参考
值----------Value----------
An object of class EBMTP. Again, for brevity, the values below represent slots which distinguish objects of class EBMTP from those of class MTP. <br>
对象类EBMTP。再次，为简单起见，下面的值代表插槽区别于类EBMTPMTP类的对象。参考
参数：<code>falsepos</code>
A matrix with rows equal to the number of hypotheses and columns the number of samples of null test statistics (B) indicating the number of guessed false positives when using the corresponding value of the observed test statistic as a cut-off. Not returned unless keep.falsepos=TRUE.
与假设和列空测试统计的样本数（B）指示的数目等于行矩阵猜测误报时，使用相应的观测试验统计值作为截断。除非keep.falsepos=TRUE没有回来。
参数：<code>truepos</code>
A matrix with rows equal to the number of hypotheses and columns the number of samples of null test statistics (B) indicating the number of guessed true positives when using the corresponding value of the observed test statistic as a cut-off. Not returned unless keep.truepos=TRUE.
与假设和列空测试统计的样本数（B）指示的数目等于行矩阵猜测真阳性时，使用相应的观测试验统计值作为截断。除非keep.truepos=TRUE没有回来。
参数：<code>errormat</code>
The matrix obtained after applying to type I error rate function closure to the matrices in falsepos, and, if applicable, truepos. Not returned unless keep.errormat=TRUE.
申请输入falsepos我的错误率函数的闭包矩阵后，矩阵获得，如果适用，truepos。除非keep.errormat=TRUE没有回来。
参数：<code>EB.h0M</code>
The sum of the local q-values obtained after density estimation. This number serves as an estimate of the proportion of true null hypotheses. Values close to one indicate situations in which type I error control may not be guaranteed by the EBMTP. When prior='EBLQV', this value is used as the prior 'pi' during evaluation of the local q-value function.
密度估计后获得当地的Q值的总和。作为一个真正的零假设的比例估计这个数字。值接近1表明，在I型误差控制，不得由EBMTP保证的情况下。当prior='EBLQV'，这个值被用来作为前“PI”在当地的Q-值函数的评价。
参数：<code>prior</code>
The numeric value of the prior 'pi' used when evaluating the local q-value function.
评估当地的Q-值函数时使用的“PI”前的数值。
参数：<code>prior.type</code>
Character string returning the value of prior in the original call to EBMTP. One of 'conservative', 'ABH', or 'EBLQV'.
字符串返回值priorEBMTP原来的呼叫。一个“保守”，“陆地棉”，或“EBLQV”。
参数：<code>lqv</code>
A numeric vector of length the number of hypotheses with the estimated local q-values used for generating guessed sets of true null hypotheses.
数字长度为向量的假设与估计本地产生的Q值猜到套真正的零假设。
参数：<code>Hsets</code>
A numeric matrix with the same dimension as nulldist, containing the Bernoulli realizations of the estimated local q-values stored in lqv which were used to partition the hypotheses into guessed sets of true and false null hypotheses at each round of (re)sampling. Not returned unless keep.Hsets=TRUE.
一个数字矩阵具有相同的尺寸为nulldist，含有lqv真假虚无假设猜到套在每个分区的假设被用来存储本地估计Q值伯努利实现的圆的（重新）抽样。除非keep.Hsets=TRUE没有回来。
作者（S）----------Author(s)----------
Houston N. Gilbert, based on the original <code>MTP</code> code written by Katherine S. Pollard
参考文献----------References----------
hypothesis testing with applications to genomics: New developments in R/Bioconductor package multtest. Journal of Statistical Software (submitted). Temporary URL: http://www.stat.berkeley.edu/~houston/JSSNullDistEBMTP.pdf.<br>
discovery rate in multiple testing with independent statistics. J. Behav. Educ. Statist. Vol 25: 60-83.<br>
procedures that control the false discovery rate. Biometrika. Vol. 93: 491-507.<br>
http://www.bepress.com/sagmb/vol4/iss1/art29/ <br>
Resampling-based empirical Bayes multiple testing procedures for controlling generalized tail probability and expected value error rates: Focus on the false discovery rate and simulation study. Biometrical Journal, 50(5):716-44. http://www.stat.berkeley.edu/~houston/BJMCPSupp/BJMCPSupp.html. <br>
graphical model selection with applications to biological networks. Technical report, U.C. Berkeley Division of Biostatistics Working Paper Series, April 2009. URL http://www.bepress.com/ucbbiostat/paper245. <br>
参见----------See Also----------
MTP, EBMTP-class, EBMTP-methods, Hsets
MTP，EBMTP-class，EBMTP-methods，Hsets
举例----------Examples----------
set.seed(99)
data<-matrix(rnorm(90),nr=9)
group<-c(rep(1,5),rep(0,5))
#EB fwer control with centered and scaled bootstrap null distribution [EB fwer控制中心和规模化的引导空分布]
#(B=100 for speed)[（二= 100的速度）]
eb.m1<-EBMTP(X=data,Y=group,alternative="less",B=100,method="common.cutoff")
print(eb.m1)
summary(eb.m1)
par(mfrow=c(2,2))
plot(eb.m1,top=9)
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 2012-2-26 07:37:52
