当 运行 GAM 中的固定效应在 R 中带有 gratia::draw() 的偏移时出错
Error when running a fixed effect in a GAM with an offset in R with gratia::draw()
我有一个如下所示的数据集：
structure(list(n = c(236896L, 73258L, 75570L, 5684L, 10242L,
2037L, 74194L, 41764L, 288115L, 6728L, 18964L, 5395L, 23192L,
12575L, 39591L, 12566L, 44458L, 126957L, 47316L, 152175L, 92913L,
81229L, 29622L, 1708L, 8526L, 52117L, 95385L, 22480L, 30521L,
51660L, 74320L, 273107L, 58L, 59686L, 77454L, 51471L, 66610L,
232321L, 53435L, 45270L), name = structure(c(9L, 9L, 6L, 5L,
2L, 5L, 6L, 9L, 6L, 4L, 4L, 4L, 4L, 2L, 9L, 2L, 6L, 1L, 4L, 6L,
4L, 9L, 2L, 5L, 3L, 4L, 6L, 2L, 9L, 2L, 4L, 4L, 7L, 9L, 1L, 6L,
6L, 6L, 8L, 2L), .Label = c("Ami", "Cho", "Fal", "For", "Ric",
"Sam", "Taw", "Tex", "Tol"), class = "factor"), change2 = c(0.0753607803884176,
-0.08058465598786, -0.00410425493512865, -0.0220964428266722,
0.0629320532004209, -0.0797306134519322, 0.0660481799732004,
-0.0572995403797303, -0.00713582946272, 0.00756646981276647,
0.032732914683994, -0.00632056690250293, 0.050358229187504, 0.0265162711945312,
0.0218803226963826, -0.0508818612242459, 0.00485925918649957,
0.0315158006542641, -0.0315622434590242, -0.0602515470219345,
-0.0409479919129347, 0.111224942380013, 0.00704490808823113,
0.0236731452544392, -0.0811686305416274, -0.0274692750452023,
0.00160881330548216, -0.0211269729894635, -0.0377625466699325,
-0.0311273993307701, -0.0118001904995042, 0.0023179680499073,
0.0263453251509878, 0.0767020512037913, -0.0113771665605732,
-0.0428469659333539, 0.0714746847470087, 0.10720066191237, 0.0153144105362596,
-0.109538998188302), Season = structure(c(2L, 1L, 4L, 3L, 3L,
1L, 4L, 3L, 2L, 4L, 4L, 4L, 1L, 4L, 3L, 1L, 3L, 4L, 1L, 1L, 3L,
2L, 2L, 2L, 3L, 3L, 3L, 1L, 3L, 4L, 1L, 2L, 3L, 2L, 2L, 3L, 4L,
2L, 3L, 2L), .Label = c("fall", "spring", "summer", "winter"), class = "factor"),
off = c(230915, 57957, 85583, 10526, 35316.6, 4851, 87287,
48226, 198700, 42050.6, 46252.8, 29974, 56566, 20959, 43175,
10385, 56997, 208126, 100672, 80516, 244507, 128730, 38470,
5177, 22435.6, 121202, 114234, 26140, 24693, 53812.6, 124281,
666114, 583, 76915, 140824.7, 91912, 78828, 219171, 95419,
33783.9)), row.names = c(NA, -40L), class = "data.frame")
我正在 运行创建一个如下所示的 GAM：
gam1<-gam(n~Season+s(change2, by=Season, k=5)+
s(name, bs="re")+
offset(log(off)),
data=data,family=nb,method="REML")
name 变量具有随机效应，Season 变量具有固定效应。
我能够使用此模型的 summary() 命令获得所有输出，但是当我尝试使用 gratia::draw() 命令显示部分效果图时，我收到以下错误：
library(gratia)
draw(gam1)
Error in eval(predvars, data, env) : object 'off' not found
In addition: Warning message:
In predict.gam(object, newdata = pred_data, type = "terms", terms = term, :
not all required variables have been supplied in newdata!
部分效果图用plot.gam函数显示，所以我想知道为什么这个模型不适用于绘制命令？如果我去掉偏移量，模型也会 运行，但是此信息对分析至关重要。为什么我不能 运行 具有偏移的固定效果？
我无法在 0.7.1 版（在 GitHub，但即将提交给 CRAN）中重现此错误，而且我认为与报告的问题相关的任何内容在 0.7 版之间都没有变化。 0（当前 CRAN 版本）和这个补丁版本。
你的数据在 tmp:
r$> gam1 <- gam(n ~ Season + s(change2, by = Season, k = 5)+
s(name, bs = "re") +
offset(log(off)),
data = tmp, family = nb, method = "REML")
r$> draw(gam1)
我得到：
我有一个如下所示的数据集：
structure(list(n = c(236896L, 73258L, 75570L, 5684L, 10242L,
2037L, 74194L, 41764L, 288115L, 6728L, 18964L, 5395L, 23192L,
12575L, 39591L, 12566L, 44458L, 126957L, 47316L, 152175L, 92913L,
81229L, 29622L, 1708L, 8526L, 52117L, 95385L, 22480L, 30521L,
51660L, 74320L, 273107L, 58L, 59686L, 77454L, 51471L, 66610L,
232321L, 53435L, 45270L), name = structure(c(9L, 9L, 6L, 5L,
2L, 5L, 6L, 9L, 6L, 4L, 4L, 4L, 4L, 2L, 9L, 2L, 6L, 1L, 4L, 6L,
4L, 9L, 2L, 5L, 3L, 4L, 6L, 2L, 9L, 2L, 4L, 4L, 7L, 9L, 1L, 6L,
6L, 6L, 8L, 2L), .Label = c("Ami", "Cho", "Fal", "For", "Ric",
"Sam", "Taw", "Tex", "Tol"), class = "factor"), change2 = c(0.0753607803884176,
-0.08058465598786, -0.00410425493512865, -0.0220964428266722,
0.0629320532004209, -0.0797306134519322, 0.0660481799732004,
-0.0572995403797303, -0.00713582946272, 0.00756646981276647,
0.032732914683994, -0.00632056690250293, 0.050358229187504, 0.0265162711945312,
0.0218803226963826, -0.0508818612242459, 0.00485925918649957,
0.0315158006542641, -0.0315622434590242, -0.0602515470219345,
-0.0409479919129347, 0.111224942380013, 0.00704490808823113,
0.0236731452544392, -0.0811686305416274, -0.0274692750452023,
0.00160881330548216, -0.0211269729894635, -0.0377625466699325,
-0.0311273993307701, -0.0118001904995042, 0.0023179680499073,
0.0263453251509878, 0.0767020512037913, -0.0113771665605732,
-0.0428469659333539, 0.0714746847470087, 0.10720066191237, 0.0153144105362596,
-0.109538998188302), Season = structure(c(2L, 1L, 4L, 3L, 3L,
1L, 4L, 3L, 2L, 4L, 4L, 4L, 1L, 4L, 3L, 1L, 3L, 4L, 1L, 1L, 3L,
2L, 2L, 2L, 3L, 3L, 3L, 1L, 3L, 4L, 1L, 2L, 3L, 2L, 2L, 3L, 4L,
2L, 3L, 2L), .Label = c("fall", "spring", "summer", "winter"), class = "factor"),
off = c(230915, 57957, 85583, 10526, 35316.6, 4851, 87287,
48226, 198700, 42050.6, 46252.8, 29974, 56566, 20959, 43175,
10385, 56997, 208126, 100672, 80516, 244507, 128730, 38470,
5177, 22435.6, 121202, 114234, 26140, 24693, 53812.6, 124281,
666114, 583, 76915, 140824.7, 91912, 78828, 219171, 95419,
33783.9)), row.names = c(NA, -40L), class = "data.frame")
我正在 运行创建一个如下所示的 GAM：
gam1<-gam(n~Season+s(change2, by=Season, k=5)+
s(name, bs="re")+
offset(log(off)),
data=data,family=nb,method="REML")
name 变量具有随机效应，Season 变量具有固定效应。
我能够使用此模型的 summary() 命令获得所有输出，但是当我尝试使用 gratia::draw() 命令显示部分效果图时，我收到以下错误：
library(gratia)
draw(gam1)
Error in eval(predvars, data, env) : object 'off' not found
In addition: Warning message:
In predict.gam(object, newdata = pred_data, type = "terms", terms = term, :
not all required variables have been supplied in newdata!
部分效果图用plot.gam函数显示，所以我想知道为什么这个模型不适用于绘制命令？如果我去掉偏移量，模型也会 运行，但是此信息对分析至关重要。为什么我不能 运行 具有偏移的固定效果？
我无法在 0.7.1 版（在 GitHub，但即将提交给 CRAN）中重现此错误，而且我认为与报告的问题相关的任何内容在 0.7 版之间都没有变化。 0（当前 CRAN 版本）和这个补丁版本。
你的数据在 tmp:
r$> gam1 <- gam(n ~ Season + s(change2, by = Season, k = 5)+
s(name, bs = "re") +
offset(log(off)),
data = tmp, family = nb, method = "REML")
r$> draw(gam1)
我得到：