huggins91(VGAM)
huggins91()所属R语言包：VGAM
Huggins (1991) Capture-recapture Model Family Function
哈金斯（1991）捕获 - 再捕获模型家庭功能
译者：生物统计家园网 机器人LoveR
描述----------Description----------
Fits a Huggins (1991) capture-recapture model to a matrix of 0s and 1s: animals sampled on several occasions and individual animals caught at least once.
适用于一个哈金斯（1991）捕获 - 再捕获模型，矩阵的“0”和“1：动物，采样过好几次，至少一次捕获的动物个体。
用法----------Usage----------
huggins91(link = "logit", earg = list(), parallel = TRUE,
iprob = NULL, eim.not.oim = TRUE)
参数----------Arguments----------
参数：link, earg, parallel, iprob
See CommonVGAMffArguments for information. The parallel argument should generally be left alone since parallelism is assumed by Huggins (1991).
见CommonVGAMffArguments的信息。 parallel参数一般应单独留下，，因为的并行假定哈金斯（1991）。
参数：eim.not.oim
Logical. If TRUE use the EIM, else the OIM.
逻辑。如果TRUE使用的EIM，否则将OIM。
Details
详细信息----------Details----------
This model operates on a response matrix of 0s and 1s. Each of at least two columns is an occasion where animals are potentially captured (e.g., a field trip), and each row is an individual animal. Capture is a 1, else a 0. Each row has at least one capture. It is well-known that animals are affected by capture, e.g., trap-shy or trap-happy. This VGAM family function attempts to allow the capture history to be modelled. This involves the use of the xij argument. Ignoring capture history effects would mean posbinomial could be used by aggregating over the sampling occasions.
此模型上操作的响应矩阵0s和1s。每个人至少有两列是一个机会，动物有可能捕获（例如，实地考察），每一行是一个动物个体。捕获是一个1，否则为0。每一行都有至少一种捕获。这是众所周知的动物都受到捕捉，例如，陷阱害羞或陷阱高兴。这VGAM家庭功能，试图让捕获的历史为蓝本。这包括使用的xij参数。忽视捕捉历史的影响将意味着posbinomial可以通过聚集在采样的场合使用。
Huggins (1991) suggests a model involving maximizing a conditional likelihood. The form of this is a numerator divided by a denominator, where the true model has part of the linear/additive predictor modelling capture history applying to the numerator only, so that part is set to zero in the denominator. The numerator of the conditional likelihood corresponds to a sequence of Bernoulli trials, with at least one success, for each animal.
哈金斯（1991）提出的模型，有条件的可能性最大化。的形式，这是一个分子除以分母，真实的模型有线性/添加剂预测模型捕捉申请的分子，所以这部分是设置为0，分母中的历史的一部分。有条件的似然率的分子对应伯努利试验的序列，与至少一个成功，对每个动物。
Unfortunately the Huggins model is too difficult to fit in this package, and one can only use the same linear/additive predictor in the numerator as the denominator. Hence this VGAM family function does not implement the model properly.
不幸的是，的哈金斯模型是太困难了适合这个包，只能使用相同的分子，分母的线性/添加剂预测。因此，这VGAM家庭功能没有实现正确。
The number of linear/additive predictors is twice the number of sampling occasions, i.e., M = 2T, say. The first two correspond to the first sampling occasion, the next two correspond to the second sampling occasion, etc. Even-numbered linear/additive predictors should correspond to what would happen if no capture had occurred (they belong to the denominator.) Odd-numbered linear/additive predictors correspond to what actually happened (they belong to the numerator.)
线性/添加剂预测因子的数目的两倍的数量的采样的场合，即，M = 2T，说。前两个对应的第一个采样的场合，接下来的两个对应的第二采样的场合，偶数线性/添加剂的预测会发生什么，如果没有捕获发生（他们属于分母。）奇编号线性/添加剂的预测究竟发生了什么（他们属于分子）。
The fitted value for column t is the tth numerator probability divided by the denominator.
拟合值的列t是t个分子除以分母的概率。
值----------Value----------
An object of class "vglmff" (see vglmff-class). The object is used by modelling functions such as vglm, and vgam.
类的一个对象"vglmff"（见vglmff-class）。该对象被用于建模功能，如vglm，vgam。
警告----------Warning ----------
This VGAM family function is experimental and does not work properly because the linear/additive predictor in the numerator and denominator must be the same. The parameter estimates of the Huggins (1991) model ought to be similar (probably in between, in some sense) to two models: Model 1 is where the capture history variable is included, Model 2 is where the capture history variable is not included. See the example below. A third model, called Model 3, allows for 'half' the capture history to be put in both numerator and denominator. This might be thought of as a compromise between Models 1 and 2, and may be useful as a crude approximation.
这VGAM家庭功能是实验性的，并不能正常工作，因为线性/对添加剂预测中的分子和分母必须是相同的。哈金斯（1991）模型的参数估计应该是相似的（可能是介于两者之间，在某种意义上）两种模型：模型1是捕捉历史的变量，模型2的捕捉历史的变量不包括。请看下面的例子。第三个模型，模型3，允许“半壁江山”捕捉历史中的分子和分母。这可能被认为作为模型1和2之间的一种折衷，并且可能是有用的，为粗制的近似。
Under- or over-flow may occur if the data is ill-conditioned.
不足或过流时可能会出现的数据是病态的。
注意----------Note----------
The weights argument of vglm need not be assigned, and the default is just a matrix of ones.
weights参数vglm不必被指定，默认是矩阵的。
This VGAM family function is currently more complicated than it needs to be, e.g., it is possible to simplify M = T, say.
是目前比较复杂得多，它需要这VGAM家庭功能，例如，它可以简化M = T，说。
（作者）----------Author(s)----------
Thomas W. Yee
参考文献----------References----------
Some practical aspects of a conditional likelihood approach to capture experiments. Biometrics, 47, 725–732.
参见----------See Also----------
vglm.control for xij, dhuggins91, rhuggins91. posbinomial.
vglm.controlxij，dhuggins91，rhuggins91。 posbinomial。
实例----------Examples----------
set.seed(123); nTimePts = 5
hdata = rhuggins91(n = 1000, nTimePts = nTimePts, pvars = 2)
# The truth: xcoeffs are c(-2, 1, 2) and capeffect = -1[真相：xcoeffs C（-2，1，2），capeffect = -1]
# Model 1 is where capture history information is used[模型图1是使用捕获历史信息]
model1 = vglm(cbind(y1, y2, y3, y4, y5) ~ x2 + Chistory,
huggins91, data = hdata, trace = TRUE,
xij = list(Chistory ~ ch0 + zch0 +
ch1 + zch1 + ch2 + zch2 +
ch3 + zch3 + ch4 + zch4 - 1),
form2 = ~ 1 + x2 + Chistory +
ch0 + ch1 + ch2 + ch3 + ch4 +
zch0 + zch1 + zch2 + zch3 + zch4)
coef(model1, matrix = TRUE) # Biased!![偏！]
summary(model1)
head(fitted(model1))
head(model.matrix(model1, type = "vlm"), 21)
head(hdata)
# Model 2 is where no capture history information is used[模式2是在没有捕捉历史信息]
model2 = vglm(cbind(y1, y2, y3, y4, y5) ~ x2,
huggins91, data = hdata, trace = TRUE)
coef(model2, matrix = TRUE) # Biased!![偏！]
summary(model2)
# Model 3 is where half the capture history is used in both[模型图3是半个捕获历史中都使用了]
# the numerator and denominator[分子和分母]
set.seed(123); nTimePts = 5
hdata2 = rhuggins91(n = 1000, nTimePts = nTimePts, pvars = 2,
double.ch = TRUE)
head(hdata2) # 2s have replaced the 1s in hdata[2中的“1”已经取代了HDATA]
model3 = vglm(cbind(y1, y2, y3, y4, y5) ~ x2 + Chistory,
huggins91, data = hdata2, trace = TRUE,
xij = list(Chistory ~ ch0 + zch0 +
ch1 + zch1 + ch2 + zch2 +
ch3 + zch3 + ch4 + zch4 - 1),
form2 = ~ 1 + x2 + Chistory +
ch0 + ch1 + ch2 + ch3 + ch4 +
zch0 + zch1 + zch2 + zch3 + zch4)
coef(model3, matrix = TRUE) # Biased!![偏！]
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 15:38:23
