带有类因子预测器的R中多项logit模型的错误

时间:2017-04-21 01:58:11

标签: r cut mlogit

我在R中使用mlogit运行多项logit模型。我的示例数据框如下所示。

  est      best_adap ele area exp
1   A  planting_date  up  169  15
2   B crop_varieties  up  441  18
3   C          shade low  461  28
4   D crop_varieties mid  324  27
5   E crop_varieties low  312  20
6   F  planting_date  up  152  30

可通过以下代码访问此数据框。

structure(list(est = structure(1:6, .Label = c("A", "B", "C", 
"D", "E", "F"), class = "factor"), best_adap = structure(c(2L, 
1L, 3L, 1L, 1L, 2L), .Label = c("crop_varieties", "planting_date", 
"shade"), class = "factor"), ele = structure(c(3L, 3L, 1L, 2L, 
1L, 3L), .Label = c("low", "mid", "up"), class = "factor"), area = c(169L, 
441L, 461L, 324L, 312L, 152L), exp = c(15L, 18L, 28L, 27L, 20L, 
30L)), .Names = c("est", "best_adap", "ele", "area", "exp"), class = "data.frame", row.names = c(NA, 
-6L))

简单模型的代码如下

#install.packages ("mlogit")
 library(mlogit)

 #Give factor Id for alternatives. "best_adap" has three alternatives

 dat["best_adap_fac"]<- NA
 dat$best_adap_fac <- factor(dat$best_adap, levels = c("planting_date"  ,"crop_varieties","shade"),labels = c(1,2,3))

 #Give factor Id for different levels of variable "ele"
 dat["ele_fac"]<- NA
 dat$ele_fac <- factor(dat$ele, levels = c("up" ,"low","mid"),labels =  c(1,2,3)) 

 #Convert experience (variable "exp") into three ranges
 dat$exp_category<-cut(dat$exp, seq(0,30,10), labels = c(1:3))


#Covert the dataset from wide to long
 dat_long<-mlogit.data(dat, varying = NULL, choice = "best_adap_fac", shape  = "wide")

#Run Multinomial logit model. We don't have any variable that varies with  the alternative but all vary only with individual.

mlogit_model1<-mlogit(best_adap_fac ~ 1| area, data=dat_long, reflevel = "3")
summary(mlogit_model1) #This runs!


mlogit_model2<-mlogit(best_adap_fac ~ 1| area + ele_fac, data=dat_long,   reflevel = "3")
summary(mlogit_model2) #This works too

mlogit_model3<-mlogit(best_adap_fac ~ 1| area + ele_fac + exp_category,  data=dat_long,    reflevel = "3") 

模型3不起作用并返回以下错误消息

Error in solve.default(H, g[!fixed]) : 
Lapack routine dgesv: system is exactly singular: U[12,12] = 0

请有人帮我找到解决方案。

谢谢

0 个答案:

没有答案