R中未标记的FitColExt的交互作用图

问题描述

在R中,我使用unmarked包安装了动态入住模型。最佳模型包括site covariateyearly-site covariategamma-formula间的一个交互项。为了解释结果,我想绘制一个交互图。但是,我不知道如何完成此操作。尤其是因为yearly-site covariate由两个向量组成(一个包含历史数据,一个包含最近数据)。

我的数据结构如下。 (不幸的是,我无法生成伪造的数据来产生有意义的模型,也许有人也可以帮忙吗?):

library (unmarked)
library (tidyverse)

M <- 40 # number of Sites
J <- 1 # num secondary sample periods
Times <- 2 # num primary sample periods

# there are two years in which 40 study plots were surveyed for the occurence of a bird species:
year1 <- c (1,1,0)
year2 <- c (0,0)
bird_territories <- array (data = c (year1,year2),dim = c (M,Times)) %>% as.matrix()

# the study plots are clustered in three regions:
region <- c (rep ("A",15),rep ("B",12),rep ("C",13))

# var_a,var_b and var_c are site-specific coefficients,not changing with time
var_a <- (year1 + 0.1) * rnorm (40)^2
var_b <- rnorm (40)^2
var_c <- 1/var_a*(year2+1)

#var_d is a site-specific coefficient that also changes with time
var_d <- data.frame ("historic" = var_b,"recent" = var_b + (as.numeric (gsub (-1,year2-year1))))


# aggregate environmental covariates
covariatesSite <- data.frame (region = region,var_a = var_a,var_b = var_b,var_c = var_c)
covariatesSiteYear <- list (var_d = var_d)

# create unmarked data frame                         
bird_data <- unmarkedMultFrame (y = bird_territories,numPrimary = Times,siteCovs = covariatesSite,yearlySiteCovs = covariatesSiteYear
                                )       

summary (bird_data)

我适合模特:

m1 <- colext(psiformula= ~ var_a + var_b,gammaformula = ~ var_c * var_d,epsilonformula = ~   1,pformula = ~ 1,data = bird_data,method="BFGS")
summary (m1) 

如前所述,我无法为该示例创建“酷”数据。但是,让我们假设摘要看起来像这样:

Call:
colext(psiformula = ~elevation + tpi2160 + coniferous_forest + 
    abandonde_pasture,gammaformula = ~sNow * tpi2160 + coniferous_forest,epsilonformula = ~1,pformula = ~1,data = gridCells_dom,method = "BFGS")

Initial (logit-scale):
                  Estimate    SE     z  P(>|z|)
(Intercept)         -0.713 0.302 -2.36 0.018104
var_a                0.617 0.177  3.49 0.000491
var_b                0.578 0.158  3.67 0.000243


Colonization (logit-scale):
                  Estimate    SE      z P(>|z|)
(Intercept)          0.211 0.513  0.412 0.68035
var_c                0.247 0.187  1.324 0.18545
var_d                0.666 0.229  2.910 0.00362

var_c:var_d         -0.652 0.304 -2.144 0.03201

Extinction (logit-scale):
 Estimate    SE      z P(>|z|)
   -0.517 0.537 -0.964   0.335

Detection (logit-scale):
 Estimate    SE    z P(>|z|)
      1.2 0.711 1.69  0.0918

AIC: 1282.268 
Number of sites: 40
optim convergence code: 0
optim iterations: 57 
Bootstrap iterations: 0 

我看到var_c和var_d之间的交互很重要。但是我无法弄清楚这些变量与影响研究鸟类物种的发生之间的关系。我认为,互动情节可能会有所帮助。但是:如何产生这样的情节?

感谢您的帮助,并为错误的假数据感到抱歉-我只是想不出如何做好更好的准备。

解决方法

暂无找到可以解决该程序问题的有效方法,小编努力寻找整理中!

如果你已经找到好的解决方法,欢迎将解决方案带上本链接一起发送给小编。

小编邮箱:dio#foxmail.com (将#修改为@)