如何在 ggplot2 中绘制来自 frair 包的 mle2 拟合？
How do I plot a mle2 fit from frair package in ggplot2?
所以我试图在 ggplot2 中绘制函数响应曲线，为此我使用了 fair 包中的 frait_fit() 和 frair_boot() 函数（https://cran.r-project.org/web/packages/frair/frair.pdf).
所以我打电话给 frair_fit:
FRAM18.fitII<- frair_fit(eaten~density, data=FRAM18,
response="rogersII",
start = list(a = 1, h = 0.1),
fixed = list(T = 3))
得到这个
> summary(FRAM18.fitII)
Length Class Mode
call 6 -none- call
x 56 -none- numeric
y 56 -none- numeric
response 1 -none- character
xvar 1 -none- character
yvar 1 -none- character
optimvars 2 -none- character
fixedvars 1 -none- character
coefficients 3 -none- numeric
sample 56 -none- numeric
fit 1 mle2 S4
frair_boot 看起来像这样：
FRAM18.bootII <- frair_boot(FRAM18.fitII, nboot=3000)
> summary(FRAM18.bootII)
Length Class Mode
call 6 -none- call
x 56 -none- numeric
y 56 -none- numeric
response 1 -none- character
xvar 1 -none- character
yvar 1 -none- character
optimvars 2 -none- character
fixedvars 1 -none- character
coefficients 3 -none- numeric
bootcoefs 9000 -none- numeric
sample 168000 -none- numeric
n_failed 1 -none- numeric
n_duplicated 1 -none- numeric
n_boot 1 -none- numeric
stratified 1 -none- logical
fit 11 boot list
我已尝试使用此问题中概述的预测和熔化函数 只得到：
> FRAM18.fitII2$mle2 <- predict(FRAM18.fitII,newdata=FRAM18.fitII2)
Error in UseMethod("predict") :
no applicable method for 'predict' applied to an object of class "c('frfit', 'list')"
与引导函数输出的结果相似。
我可以使用
创建图表
plot(FRAM18.fitII$x,FRAM18.fitII$y, pch=20, col='skyblue2', xlim=c(0,300))
lines(FRAM18.fitII, col='skyblue3')
但我想要更具视觉吸引力的东西。
frair 包告诉我它使用 mle2 进行计算，所以我想知道是否还有办法将其输出仍然输入 ggplot2
我的数据是这样的：
density
eaten
180
40
180
4
160
70
180
55
100
13
50
16
25
4
15
15
140
46
160
22
25
25
50
0
25
18
160
11
100
6
50
50
100
75
15
9
15
15
140
138
140
140
加载包和数据，进行初始拟合：
library(frair)
library(emdbook)
data(ReedfrogFuncresp)
newfit <- frair_fit(Killed~Initial, data=ReedfrogFuncresp,
response="rogersII",
start = list(a = 1, h = 0.1),
fixed = list(T = 1))
这是拟合模型的 predict 方法，它使用 bootstrap 对象（如果提供）来构造置信区间：
#' @param object fitted frair model
#' @param newdata specified densities for prediction; data frame
#' with density column name matching that of original model (\code{object$xvar})
#' @param boot output of \code{frair_boot}
#' @param quantiles for confidence intervals
predict.frfit <- function(object, newdata = NULL, boot = NULL, quantiles = c(0.025, 0.975)) {
fitfun <- get(object$response, pos = "package:frair")
if (!is.null(newdata)) {
newx <- newdata[[object$xvar]]
} else {
newx <- object$x
}
fitted <- fitfun(newx, as.list(object$coefficients))
if (is.null(boot)) return(fitted)
nms <- names(object$coefficients)
bootvars <- boot$fit$t[,names(boot$fit$t0) %in% nms]
colnames(bootvars) <- nms
bootres <- apply(bootvars, 1,
function(b) fitfun(newx, as.list(b)))
envelope <- t(apply(bootres, 1, quantile, quantiles))
ret <- data.frame(fitted, envelope)
names(ret) <- c(object$yvar, "lwr", "upr")
return(ret)
}
library(ggplot2)
pframe0 <- with(ReedfrogFuncresp,
data.frame(Initial = seq(0, max(Initial), length = 101)))
## basic (predicted value only)
pframe <- data.frame(pframe0, Killed= predict(newfit, newdata=pframe0))
## with confidence intervals
b <- frair_boot(newfit, ncores=3, nboot=250)
pframe_b <- data.frame(pframe0, predict(newfit, newdata=pframe0, boot = b))
ggplot(ReedfrogFuncresp, aes(Initial, Killed)) +
geom_point() +
geom_line(data = pframe_b) +
geom_ribbon(data = pframe_b, aes(ymin=lwr, ymax=upr),
colour = NA, alpha = 0.4)
所以我试图在 ggplot2 中绘制函数响应曲线，为此我使用了 fair 包中的 frait_fit() 和 frair_boot() 函数（https://cran.r-project.org/web/packages/frair/frair.pdf). 所以我打电话给 frair_fit:
FRAM18.fitII<- frair_fit(eaten~density, data=FRAM18,
response="rogersII",
start = list(a = 1, h = 0.1),
fixed = list(T = 3))
得到这个
> summary(FRAM18.fitII)
Length Class Mode
call 6 -none- call
x 56 -none- numeric
y 56 -none- numeric
response 1 -none- character
xvar 1 -none- character
yvar 1 -none- character
optimvars 2 -none- character
fixedvars 1 -none- character
coefficients 3 -none- numeric
sample 56 -none- numeric
fit 1 mle2 S4
frair_boot 看起来像这样：
FRAM18.bootII <- frair_boot(FRAM18.fitII, nboot=3000)
> summary(FRAM18.bootII)
Length Class Mode
call 6 -none- call
x 56 -none- numeric
y 56 -none- numeric
response 1 -none- character
xvar 1 -none- character
yvar 1 -none- character
optimvars 2 -none- character
fixedvars 1 -none- character
coefficients 3 -none- numeric
bootcoefs 9000 -none- numeric
sample 168000 -none- numeric
n_failed 1 -none- numeric
n_duplicated 1 -none- numeric
n_boot 1 -none- numeric
stratified 1 -none- logical
fit 11 boot list
我已尝试使用此问题中概述的预测和熔化函数
> FRAM18.fitII2$mle2 <- predict(FRAM18.fitII,newdata=FRAM18.fitII2)
Error in UseMethod("predict") :
no applicable method for 'predict' applied to an object of class "c('frfit', 'list')"
与引导函数输出的结果相似。
我可以使用
创建图表plot(FRAM18.fitII$x,FRAM18.fitII$y, pch=20, col='skyblue2', xlim=c(0,300))
lines(FRAM18.fitII, col='skyblue3')
但我想要更具视觉吸引力的东西。
frair 包告诉我它使用 mle2 进行计算，所以我想知道是否还有办法将其输出仍然输入 ggplot2
我的数据是这样的：
| density | eaten |
|---|---|
| 180 | 40 |
| 180 | 4 |
| 160 | 70 |
| 180 | 55 |
| 100 | 13 |
| 50 | 16 |
| 25 | 4 |
| 15 | 15 |
| 140 | 46 |
| 160 | 22 |
| 25 | 25 |
| 50 | 0 |
| 25 | 18 |
| 160 | 11 |
| 100 | 6 |
| 50 | 50 |
| 100 | 75 |
| 15 | 9 |
| 15 | 15 |
| 140 | 138 |
| 140 | 140 |
加载包和数据，进行初始拟合：
library(frair)
library(emdbook)
data(ReedfrogFuncresp)
newfit <- frair_fit(Killed~Initial, data=ReedfrogFuncresp,
response="rogersII",
start = list(a = 1, h = 0.1),
fixed = list(T = 1))
这是拟合模型的 predict 方法，它使用 bootstrap 对象（如果提供）来构造置信区间：
#' @param object fitted frair model
#' @param newdata specified densities for prediction; data frame
#' with density column name matching that of original model (\code{object$xvar})
#' @param boot output of \code{frair_boot}
#' @param quantiles for confidence intervals
predict.frfit <- function(object, newdata = NULL, boot = NULL, quantiles = c(0.025, 0.975)) {
fitfun <- get(object$response, pos = "package:frair")
if (!is.null(newdata)) {
newx <- newdata[[object$xvar]]
} else {
newx <- object$x
}
fitted <- fitfun(newx, as.list(object$coefficients))
if (is.null(boot)) return(fitted)
nms <- names(object$coefficients)
bootvars <- boot$fit$t[,names(boot$fit$t0) %in% nms]
colnames(bootvars) <- nms
bootres <- apply(bootvars, 1,
function(b) fitfun(newx, as.list(b)))
envelope <- t(apply(bootres, 1, quantile, quantiles))
ret <- data.frame(fitted, envelope)
names(ret) <- c(object$yvar, "lwr", "upr")
return(ret)
}
library(ggplot2)
pframe0 <- with(ReedfrogFuncresp,
data.frame(Initial = seq(0, max(Initial), length = 101)))
## basic (predicted value only)
pframe <- data.frame(pframe0, Killed= predict(newfit, newdata=pframe0))
## with confidence intervals
b <- frair_boot(newfit, ncores=3, nboot=250)
pframe_b <- data.frame(pframe0, predict(newfit, newdata=pframe0, boot = b))
ggplot(ReedfrogFuncresp, aes(Initial, Killed)) +
geom_point() +
geom_line(data = pframe_b) +
geom_ribbon(data = pframe_b, aes(ymin=lwr, ymax=upr),
colour = NA, alpha = 0.4)