我想要一个混合效果模型,它允许我考虑不同地理位置之间的不等差异。具体来说,我希望将response
预测为具有X
的固定效果geo
的函数作为随机效应。
以下是数据的样子:
X response geo
1 4 5.521461 other
2 4 5.164786 other
3 4 5.164786 other
4 6 3.401197 other
5 5 4.867534 other
6 4 5.010635 other
地理列的唯一值:
[1] "other" "Atlanta-Sandy Springs-Marietta, GA" "Chicago-Naperville-Joliet, IL-IN-WI" "Dallas-Fort Worth-Arlington, TX"
[5] "Houston-Sugar Land-Baytown, TX" "Los Angeles-Long Beach-Santa Ana, CA" "Miami-Fort Lauderdale-Pompano Beach, FL" "Phoenix-Mesa-Glendale, AZ"
以下是我尝试过的模型:
> lme0 <- lme(response ~ factor(predictor) , random = ~1|factor(geo), data = HC_hired)
> summary(lme0)
Linear mixed-effects model fit by REML
Data: HC_hired
AIC BIC logLik
54770.69 54836.3 -27377.34
Random effects:
Formula: ~1 | factor(geo)
(Intercept) Residual
StdDev: 0.08689381 0.66802
Fixed effects: response ~ factor(predictor)
Value Std.Error DF t-value p-value
(Intercept) 4.255531 0.04410213 26918 96.49264 0.0000
factor(predictor)2 0.022986 0.03336742 26918 0.68889 0.4909
factor(predictor)3 0.166341 0.03221410 26918 5.16361 0.0000
factor(predictor)4 0.299172 0.03194177 26918 9.36618 0.0000
factor(predictor)5 0.378645 0.03249053 26918 11.65402 0.0000
factor(predictor)6 0.472583 0.03664732 26918 12.89543 0.0000
Correlation:
(Intr) fct()2 fct()3 fct()4 fct()5
factor(predictor)2 -0.660
factor(predictor)3 -0.683 0.903
factor(predictor)4 -0.689 0.912 0.945
factor(predictor)5 -0.679 0.897 0.930 0.940
factor(predictor)6 -0.603 0.795 0.824 0.832 0.819
Standardized Within-Group Residuals:
Min Q1 Med Q3 Max
-4.7047458 -0.3424262 0.1883132 0.7045260 2.1949313
Number of Observations: 26931
Number of Groups: 8
我的问题是输出没有为每个级别的地理位置指定随机效果。这样做的正确型号规范是什么?我没有运气,尝试过很多公式的排列。对整个过程的任何评论也是受欢迎的。非常感谢提前!
对评论的回应(强迫地理因素不会改变输出):
HC_hired $ geo&lt; - as.factor(HC_hired $ geo) lme0&lt; - lme(响应〜因子(预测变量),随机= ~1 |因子(地理),数据= HC_hired) 摘要(lme0)