model.average.marklist(RMark)
model.average.marklist()所属R语言包：RMark
Compute model averaged estimates of real parameters
平均计算模型的实际参数的估计
译者：生物统计家园网 机器人LoveR
描述----------Description----------
Computes model averaged estimates and standard errors of real parameters for a list of models with a model.table constructed from collect.models. It can also optionally compute the var-cov matrix of the averaged parameters and their confidence intervals by transforming with the link functions, setting normal confidence intervals on the transformed values and then back-transforming for the real estimates.
与model.tablecollect.models构建计算模型的平均估计值和标准差的实际参数列表模式。它也可以选择计算的平均参数及其置信区间的VAR-COV矩阵的链接功能，通过改变正常的置信区间设置转换后的值，然后再转化为实际的估计。
用法----------Usage----------
## S3 method for class 'marklist'
model.average(x, parameter, data, vcv,
drop=TRUE, indices=NULL, revised=TRUE,...)
参数----------Arguments----------
参数：x
a list of mark model results and a model.table constructed by collect.models
构建了collect.models的标记模型结果和model.table的的一个列表
参数：parameter
name of model parameter (e.g., "Phi" for CJS models); if left NULL all real parameters are averaged
如果左NULL实际参数的平均值名称的模型参数（例如，“披”CJS模型）;
参数：data
dataframe with covariate values that are averaged for estimates
数据框的协变量值的平均值估算
参数：vcv
logical; if TRUE then the var-cov matrix and confidence intervals are computed
逻辑;如果是TRUE，则VAR-COV矩阵和置信区间的计算
参数：drop
if TRUE, models with any non-positive variance for betas are dropped
如果为TRUE，任何非正方差模型的贝他被丢弃
参数：indices
a vector of parameter indices from the all-different PIM formulation of the parameter estimates that should be presented. This argument only works if the parameter argument = NULL. The primary purpose of the argument is to trim the list of parameters in computing a vcv matrix of the real parameters which can get too big to be computed with the available memory
一个向量从不同的PIM制定的参数估计，应提交的参数指标。如果此参数只能用于参数变量= NULL。的说法是修剪的主要目的列表中的参数计算VCV矩阵，可以得到真正的参数过大而无法计算的可用内存
参数：revised
if TRUE, uses revised variance formula (eq 6.12 from Burnham and Anderson) for model averaged estimates and eq 6.11 when FALSE
如果为TRUE，使用经修订的方差公式（6.12伯纳姆和安德森式）模型的平均估计和均衡器6.11 FALSE
参数：...
additional arguments passed to specific functions
额外的参数传递给特定的功能
Details
详细信息----------Details----------
If there are any models in the model.list which do not have any output or results they are dropped. If any have non-positive variances for the betas and drop=TRUE, then the model is reported and dropped from the model averaging. The weights are renormalized for the remaining models that are not dropped before they are averaged.
如果有任何车型在没有任何输出或结果，他们都将被丢弃model.list。如果有任何不积极的测试版的差异和drop=TRUE，然后模型报告和模型平均下降至。其余的型号，没有下降之前，平均为归一化的权重。
If parameter=NULL, all real parameters are model averaged but the design data is not copied over because it can vary by the type of parameter. It is only necessary to model average all parameters at once to get covariances of model averaged parameters of differing types.
如果parameter=NULL，所有的实际参数模型的平均值，但在设计数据是不可复制的，因为它可以通过不同的参数类型。这是只需要模型的所有参数的平均值，一旦得到平均协方差模型参数的不同类型。
If data=NULL, the average covariate values are used for any models using covariates. Note that this will only work with models created after v1.5.0 such that average covariate values are stored in each model object.
如果data=NULL，平均协变量的值用于任何使用的协变量的模型。请注意，这只会工作后V1.5.0等，平均协变量的值存储在每一个模型对象创建的模型。
值----------Value----------
If vcv=FALSE, the return value is a dataframe of model averaged estimates and standard errors for a particular type of real parameter (e.g., Phi). The design data are appended to the dataframe to enable subsettting of the estimates based on features of the design data such as age, time, cohort and grouping variables.
如果VCV = FALSE，则返回值是一个数据框模型的估计和标准误差平均为特定类型的实际参数（例如，皮皮）。设计数据被附加到的数据框，以使subsettting的基于功能的设计数据，如年龄，时间，队列和分组变量的估计。
If vcv=TRUE, confidence interval (lcl,ucl) limits are added to the dataframe which is contained in a list with the var-cov matrix.
如果VCV = TRUE，置信区间（LCL，UCL）的限制将被添加到数据框，其中包含在一个列表中的变种病毒矩阵的。
（作者）----------Author(s)----------
Jeff Laake
参考文献----------References----------
and Multimodel Inference: A Practical Information-Theoretic Approach, Second edition. Springer, New York.
参见----------See Also----------
collect.models, covariate.predictions, model.table, compute.links.from.reals, model.average.list
collect.models，covariate.predictions，model.table，compute.links.from.reals，model.average.list
实例----------Examples----------
data(dipper)
run.dipper=function()
{
#[]
# Process data[过程数据]
#[]
dipper.processed=process.data(dipper,groups=("sex"))
#[]
# Create default design data[创建默认的设计数据]
#[]
dipper.ddl=make.design.data(dipper.processed)
#[]
# Add Flood covariates for Phi and p that have different values[加入洪水的协变量披和p有不同的价值观]
#[]
dipper.ddl$Phi$Flood=0
dipper.ddl$Phi$Flood[dipper.ddl$Phi$time==2 | dipper.ddl$Phi$time==3]=1
dipper.ddl$p$Flood=0
dipper.ddl$p$Flood[dipper.ddl$p$time==3]=1
#[]
# Define range of models for Phi[定义范围内的车型披]
#[]
Phi.dot=list(formula=~1)
Phi.time=list(formula=~time)
Phi.sex=list(formula=~sex)
Phi.sextime=list(formula=~sex+time)
Phi.sex.time=list(formula=~sex*time)
Phi.Flood=list(formula=~Flood)
#[]
# Define range of models for p[定义范围内的车型为p]
#[]
p.dot=list(formula=~1)
p.time=list(formula=~time)
p.sex=list(formula=~sex)
p.sextime=list(formula=~sex+time)
p.sex.time=list(formula=~sex*time)
p.Flood=list(formula=~Flood)
#[]
# Collect pairings of models[收集配对的模型]
#[]
cml=create.model.list("CJS")
#[]
# Run and return the list of models[运行并返回的型号列表]
#[]
return(mark.wrapper(cml,data=dipper.processed,ddl=dipper.ddl))
}
dipper.results=run.dipper()
Phi.estimates=model.average(dipper.results,"Phi",vcv=TRUE)
p.estimates=model.average(dipper.results,"p",vcv=TRUE)
run.dipper=function()
{
data(dipper)
dipper$nsex=as.numeric(dipper$sex)-1
#NOTE: This generates random valules for the weights so the answers using[注：为权重，产生随机valules，这样的答案使用]
# ~weight will vary[~体重会有所不同]
dipper$weight=rnorm(294)
mod1=mark(dipper,groups="sex",
model.parameters=list(Phi=list(formula=~sex+weight)))
mod2=mark(dipper,groups="sex",
model.parameters=list(Phi=list(formula=~sex)))
mod3=mark(dipper,groups="sex",
model.parameters=list(Phi=list(formula=~weight)))
mod4=mark(dipper,groups="sex",
model.parameters=list(Phi=list(formula=~1)))
dipper.list=collect.models()
return(dipper.list)
}
dipper.results=run.dipper()
real.averages=model.average(dipper.results,vcv=TRUE)
# get model averaged estimates for all parameters and use average[模型的所有参数的平均估计，并使用平均]
# covariate values in models with covariates[在模型中的协变量的协变量值]
real.averages$estimates
# get model averaged estimates for Phi using a value of 2 for weight[使用重量值2，模型的平均估计披]
model.average(dipper.results,"Phi",
data=data.frame(weight=2),vcv=FALSE)
# what you can't do yet is use different covariate values for[你不能这样做，但使用不同的协变量值]
# different groups to get covariances of estimates based on different[不同的群体协方差估计，根据不同的]
# covariate values; for example, you can get average survival of females[协变量值，例如，你可以得到女性平均生存期]
# at average female weight and average survival of males at average[女性平均体重和平均生存期的男性平均]
# male weight in separate calls to model.average but not in the same call[男性在单独的调用model.average重量，但不相同的呼叫]
# to get covariances; however, if you standardized weight by group[协方差;但是，如果你组标准重量]
# (ie stdwt = weight - groupmean) then using 0 for the covariate value would give[（即stdwt =体重 - groupmean），那么协变量的值使用0]
# the model averaged Phi by group at the average group weights and its[该模型的平均披组的平均权重和]
# covariance. You can do the above for[协方差。你可以做以上]
# a single model with find.covariates/fill.covariates.[一个单一模式与find.covariates / fill.covariates。]
# get model averaged estimates of first Phi(1) and first p(43) and v-c matrix[浏览模型平均第一披（1）和第一p（43）和vc矩阵的估计]
model.average(dipper.results,vcv=TRUE,indices=c(1,43))
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 2012-9-26 23:45:02
