dirmultinomial(VGAM)
dirmultinomial()所属R语言包：VGAM
Fitting a Dirichlet-Multinomial Distribution
一类Dirichlet多项分布拟合
译者：生物统计家园网 机器人LoveR
描述----------Description----------
Fits a Dirichlet-multinomial distribution to a matrix response.
适合的Dirichlet多项式分布矩阵的反应。
----------Usage----------
dirmultinomial(lphi="logit", ephi = list(), iphi = 0.10,
parallel= FALSE, zero="M")
参数----------Arguments----------
参数：lphi
Link function applied to the phi parameter, which lies in the open unit interval (0,1). See Links for more choices.
Link功能的phi参数，而在于在打开的单元间隔(0,1)。见Links更多的选择。
参数：ephi
List. Extra argument for lphi. See earg in Links for general information.
列表。额外参数lphi。见earg中Links的一般信息。
参数：iphi
Numeric. Initial value for phi. Must be in the open unit interval (0,1). If a failure to converge occurs try assigning this argument a different value.
数字。初始值phi。必须在开单位的时间间隔(0,1)。如果收敛失败时尝试不同的值指定此参数。
参数：parallel
A logical (formula not allowed here) indicating whether the probabilities pi_1,…,pi_{M-1} are to be equal via equal coefficients. Note pi_M will generally be different from the other probabilities. Setting parallel=TRUE will only work if you also set zero=NULL because of interference between these arguments (with respect to the intercept term).
一个逻辑（这里不允许公式）的概率是否pi_1,…,pi_{M-1}是要通过平等系数相等。注pi_M一般会从其他的概率不同。设置parallel=TRUE只会工作，如果你还设置zero=NULL因为这些参数间的相互干扰（带截距项）。
参数：zero
An integer-valued vector specifying which linear/additive predictors are modelled as intercepts only. The values must be from the set \{1,2,…,M\}. If the character "M" then this means the numerical value M, which corresponds to linear/additive predictor associated with phi. Setting zero=NULL means none of the values from the set \{1,2,…,M\}.
指定一个整数值向量线性/添加剂的预测模型仅作为拦截。这些值必须是从集合\{1,2,…,M\}。如果字符"M"然后这意味着数值M，它对应于与phi线性/添加剂预测器。设置zero=NULL是指没有值的集合\{1,2,…,M\}。
Details
详细信息----------Details----------
The Dirichlet-multinomial distribution arises from a multinomial distribution where the probability parameters are not constant but are generated from a multivariate distribution called the Dirichlet distribution. The Dirichlet-multinomial distribution has probability function
产生的Dirichlet多项式分布多项分布的概率是不恒定的，但产生的一个多变量的分布称为Dirichlet分布。的Dirichlet多项式分布概率函数
where phi is the over-dispersion parameter and N_* = y_1+\cdots+y_M. Here, C_b^a means “a choose b” and refers to combinations (see choose). The above formula applies to each row of the matrix response. In this VGAM family function the first M-1 linear/additive predictors correspond to the first M-1 probabilities via
phi是过度分散参数和N_* = y_1+\cdots+y_M。在这里，C_b^a的意思是“a选择b”是指组合（见choose）。适用于上述式的矩阵的每一行响应。在这VGAM家庭功能的第一个M-1线性/添加剂的预测相对应的第一个M-1概率通过
where eta_j is the jth linear/additive predictor (eta_M=0 by definition for P[Y=M] but not for phi) and j=1,…,M-1. The Mth linear/additive predictor corresponds to lphi applied to phi.
Note that E(Y_j) = N_* pi_j but the probabilities (returned as the fitted values) pi_j are bundled together as a M-column matrix. The quantities N_* are returned as the prior weights.
请注意这E(Y_j) = N_* pi_j但概率（返回的拟合值）pi_j是捆绑在一起作为一个M列的矩阵。的数量N_*返回作为事先的权重。
The beta-binomial distribution is a special case of the Dirichlet-multinomial distribution when M=2; see betabinomial. It is easy to show that the first shape parameter of the beta distribution is shape1=pi*(1/phi-1) and the second shape parameter is shape2=(1-pi)*(1/phi-1). Also, phi=1/(1+shape1+shape2), which is known as the intra-cluster correlation coefficient.
β-二项分布是一种特殊的Dirichlet多项式分布的情况下，当M=2;betabinomial。这是很容易证明的第一个Beta分布的形状参数shape1=pi*(1/phi-1)和第二个形状参数是shape2=(1-pi)*(1/phi-1)的。另外phi=1/(1+shape1+shape2)作为聚类内的相关系数，这是众所周知的。
值----------Value----------
An object of class "vglmff" (see vglmff-class). The object is used by modelling functions such as vglm, rrvglm and vgam.
类的一个对象"vglmff"（见vglmff-class）。该对象被用于建模功能，如vglm，rrvglm和vgam。
If the model is an intercept-only model then @misc (which is a list) has a component called shape which is a vector with the M values pi_j * (1/phi-1).
如果模型是一个仅截距模型然后@misc（这是一个列表）有一个组件称为shape是一个向量与M值pi_j * (1/phi-1)。
警告----------Warning ----------
This VGAM family function is prone to numerical problems, especially when there are covariates.
这VGAM家庭功能是容易出现的数值的问题，尤其是当有协变量。
注意----------Note----------
The response can be a matrix of non-negative integers, or else a matrix of sample proportions and the total number of counts in each row specified using the weights argument. This dual input option is similar to multinomial.
响应可以是一个矩阵的非负整数，否则的矩阵的样本比例和总数在每行中使用weights论点指定的计数。这双输入选项类似multinomial。
To fit a "parallel" model with the phi parameter being an intercept-only you will need to use the constraints argument.
为了适应一个水货模型phi参数仅截距，您将需要使用constraints参数。
Currently, Fisher scoring is implemented. To compute the expected information matrix a for loop is used; this may be very slow when the counts are large. Additionally, convergence may be slower than usual due to round-off error when computing the expected information matrices.
目前，费舍尔得分的实施。为了计算预期的信息矩阵的一个for循环使用，这可能是非常缓慢的，当数大。此外，收敛速度比平常慢时，由于舍入误差计算预期的信息矩阵。
（作者）----------Author(s)----------
Thomas W. Yee
参考文献----------References----------
Fisher information matrix of the Dirichlet-multinomial distribution. Biometrical Journal, 47, 230–236.
Overdispersion in allelic counts and <code>θ</code>-correction in forensic genetics. Theoretical Population Biology, 78, 200–210.
参见----------See Also----------
dirmul.old, betabinomial, betabinomial.ab, dirichlet, multinomial.
dirmul.old，betabinomial，betabinomial.ab，dirichlet，multinomial。
实例----------Examples----------
n <- 10; M <- 5
y <- round(matrix(runif(n*M)*10, n, M)) # Integer counts[整型数]
fit <- vglm(y ~ 1, dirmultinomial, trace = TRUE)
head(fitted(fit))
fit@y # Sample proportions[样本比例]
weights(fit, type = "prior", matrix = FALSE) # Total counts per row[总计数，每行]
x <- runif(n)
fit <- vglm(y ~ x, dirmultinomial, trace = TRUE)
## Not run: [＃不运行：]
Coef(fit) # This does not work[这并不工作，]
## End(Not run)[＃（不执行）]
coef(fit, matrix = TRUE)
(sfit <- summary(fit))
vcov(sfit)
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 15:32:27
