從模型中提取指數系數 MuMIn::dredge

Question

我正在嘗試對來自MuMIn::dredge dredge object 的多個泊松模型的系數和標准誤差取冪，以與texreg::screenreg screenreg 一起使用。 在一個 model 的情況下，我們可以這樣做：

library(MuMIn)
library(texreg)
data(Cement)
m1 <- glm(X1 ~ y, data = Cement, na.action = "na.fail", family = poisson())
screenreg(m1)
# ==========================
#                 Model 1   
# --------------------------
# (Intercept)      -2.39 ** 
#                  (0.88)   
# y                 0.04 ***
#                  (0.01)   
# --------------------------
# AIC              76.03    
# BIC              77.16    
# Log Likelihood  -36.02    
# Deviance         26.76    
# Num. obs.        13       
# ==========================
# *** p < 0.001, ** p < 0.01, * p < 0.05



#EXP VERSION:
tr <- texreg::extract(m1)
screenreg(m1, override.coef = exp(tr@coef), override.se = exp(tr@se))
# ==========================
#                 Model 1   
# --------------------------
# (Intercept)       0.09 ** 
#                  (2.41)   
# y                 1.05 ***
#                  (1.01)   
# --------------------------
# AIC              76.03    
# BIC              77.16    
# Log Likelihood  -36.02    
# Deviance         26.76    
# Num. obs.        13       
# ==========================
# *** p < 0.001, ** p < 0.01, * p < 0.05
# Warning message:
# In override(models, override.coef, override.se, override.pvalues,  :
#   Standard errors were provided using 'override.se', but p-values were not replaced!

現在，如果我們使用dredge查看所有組合並得到沒有指數值的 output：

  fm1 <- glm(X1 ~ ., data = Cement, na.action = "na.fail", family = poisson())
dd <- dredge(fm1)
#shorten output
dd <- dd[1:4, ]
screenreg(dd)
# =============================================================
#                 Model 1    Model 2     Model 3     Model 4   
# -------------------------------------------------------------
# (Intercept)       0.98       3.95 ***    7.74 **     3.45 ***
#                  (1.02)     (0.33)      (2.07)      (0.23)   
# X3               -0.12 **   -0.16 ***   -0.15 ***   -0.15 ***
#                  (0.03)     (0.03)      (0.03)      (0.03)   
# y                 0.02 *                                     
#                  (0.01)                                      
# X4                          -0.02       -0.06 *              
#                             (0.01)      (0.03)               
# X2                                      -0.05                
#                                         (0.03)               
# -------------------------------------------------------------
# Log Likelihood  -24.98     -25.95      -24.22      -28.48    
# AICc             58.62      60.56       61.44       62.16    
# Delta             0.00       1.94        2.82        3.55    
# Weight            0.56       0.21        0.14        0.09    
# Num. obs.        13         13          13          13       
# =============================================================
# *** p < 0.001, ** p < 0.01, * p < 0.05

現在我不確定如何獲得dd object 的所有指數值，我在想這樣的事情：

tr <- texreg::extract(dd)
screenreg(dd, override.coef = exp(tr@coef), override.se = exp(tr@se))

這行不通，我想有一個快速的解決方法嗎？

謝謝

Answer 1

“平均” texreg的 texreg extract方法返回一個texreg對象列表，每個 model 有一個元素。您需要將@coef和@se應用於每個元素。 例如：

library(MuMIn)
library(texreg)

dd <- dredge(lm(y ~ X1 * X2, Cement, na.action = na.fail))
tr <- extract(dd)

screenreg(x, override.coef = lapply(lapply(tr, slot, 'coef'), exp))