varbvsbinoptimize(varbvs)
varbvsbinoptimize()所属R语言包:varbvs
Coordinate ascent for variational approximation to Bayesian
协调上升的变分近似贝叶斯
译者:生物统计家园网 机器人LoveR
描述----------Description----------
varbvsbinoptimize implements the fully-factorized variational approximation for Bayesian variable selection in logistic regression. It finds the "best" fully-factorized variational approximation to the posterior distribution of the coefficients in a logistic regression model of a binary outcome (e.g. disease status in a case-control study), with spike and slab priors on the coefficients. By "best", we mean the approximating distribution that locally minimizes the Kullback-Leibler divergence between the approximating distribution and the exact posterior.
varbvsbinoptimize实现的完全工厂化变的近似贝叶斯变量选择在logistic回归分析。找到“最好的”完全工厂化的变分近似的后验分布的二进制结果(如疾病状态的病例对照研究)logistic回归模型的系数,与穗板先验的系数。 “最好”,我们指的是近似的分布,局部最小化的Kullback-Leibler散度之间的近似分布和精确的后。
varbvsbinupdate runs a single iteration of the coordinate ascent updates to maximize the variational lower bound or, equivalently,
varbvsbinupdate运行的单次迭代的坐标上升更新最大化的变分下界,或等价地,
用法----------Usage----------
varbvsbinoptimize(X, y, sa, logodds, alpha0 = NULL, mu0 = NULL,
eta0 = NULL, fixed.eta = FALSE, verbose = TRUE)
varbvsbinupdate(X, sa, logodds, stats, alpha0, mu0, Xr0, S)
参数----------Arguments----------
参数:X
Matrix of observations about the variables (or features). It has n rows and p columns, where n is the number of samples, and p is the number of variables.
矩阵观测的变量(或功能)。 n行p列,其中n的样本数,和p是变量的数目。
参数:sa
Prior variance of the regression coefficients.
之前的回归系数的方差。
参数:logodds
Prior log-odds of inclusion for each variable. It is equal to logodds = log(q/(1-q)), where q is the prior probability that each variable is included in the linear model of Y. It may either be a scalar, in which case all the variables have the same prior inclusion probability, or it may be a vector of length p.
在此之前登录赔率包含的每个变量。这是等于logodds = log(q/(1-q)),q是先验概率,每个变量的线性模型Y的,它可以是一个标量,在这种情况下,所有的变量具有相同的前列入概率,或者它可以是一个矢量的长度p。
参数:alpha0
Initial variational estimate of posterior inclusion probabilities. It is a vector of length p. If alpha0 = NULL, the variational parameters are initialized at random.
初始变的估计后列入概率。它是一个向量的长度p。如果alpha0 = NULL,变参数的随机初始化。
参数:mu0
Initial variational estimate of posterior mean coefficients. It is a vector of length p. If mu0 = NULL, the variational parameters are randomly initialized.
初始变后平均系数的估计。它是一个向量的长度p。如果mu0 = NULL,变参数的随机初始化。
参数:y
Vector of observations about the binary outcome. It is a vector of length n.
矢量观测的二进制结果。它是一个向量的长度n。
参数:eta0
Initial free parameters specifying the variational lower bound on the logistic regression factors. If eta0 = NULL, the free parameters are initialized to a vector of ones.
首次自由参数指定的logistic回归因素的变化下限。如果eta0 = NULL,自由的参数被初始化为一个向量的。
参数:fixed.eta
Set fixed.eta = TRUE to prevent ETA from being updated.
值设定fixed.eta = TRUE的,防止ETA被更新。
参数:verbose
Set verbose = FALSE to turn off reporting the algorithm's progress.
设置verbose = FALSE关闭算法的进展情况报告。
参数:stats
This is the return value of updatestats(X,y,eta). See updatestats for more information on this argument.
这是返回值updatestats(X,y,eta)。见updatestats的更多信息,对这样的说法。
参数:Xr0
Equal to X %*% (alpha0*mu0).
等于X %*% (alpha0*mu0)。
参数:S
Order in which the coordinates are updated. It is a vector of any length. Each entry of S must be an integer between 1 and p.
顺序的坐标被更新。它是一个任意长度的向量。每个条目S必须是1之间的整数和p。
Details
详细信息----------Details----------
alpha, mu and s are the parameters of the variational approximation and, equivalently, variational estimates of posterior quantites: under the variational approximation, the ith regression coefficient is normal with probability alpha[i]; mu[i] and s[i] are the mean and variance of the coefficient given that it is included in the model. alpha, mu and s are always column vectors of length p.
alpha,mu和s的变分近似的参数,并等价,变分后分位数的估计:变分近似下,第i个回归系数的概率是正常<X >alpha[i]和mu[i]是,它是包含在模型中的系数的均值和方差。 s[i],alpha和mu总是列向量的长度s。
Unlike varbvsoptimize, y and X must not be centered. Instead, we will account for the intercept as we update the variational approximation.
不同varbvsoptimize,y和X不集中。相反,我们将占到拦截我们更新了变分近似。
Also note that the residual variance parameter sigma is not needed to model a binary trait.
还要注意的是不需要的剩余方差参数sigma模型二元性状。
The computational complexity of running varbvsbinupdate is O(n*length(S)). For efficient computation, most of the work is done by varbvsbinupdateR, a function implemented in C. The call to
计算复杂度的运行varbvsbinupdate是O(n*length(S))。为了有效地进行计算,大部分的工作是由的varbvsbinupdateR,用C实现的功能,调用
值----------Value----------
varbvsbinoptimize returns a list containing five components: variational parameters alpha, mu and s, the vector of free parameters eta, and the variational estimate of the marginal log-likelihood lnZ.
varbvsbinoptimize返回一个列表,其中包含五部分组成:变参数alpha,mu和s,自由参数向量eta,和变估计的边际对数似然lnZ。
varbvsbinupdate returns a list containing three components: the updated variational parameters alpha and mu, and the
varbvsbinupdate返回一个列表,其中包含三个组成部分:更新的变参数alpha和mu,和
(作者)----------Author(s)----------
Peter Carbonetto
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。
|