堆栈内存溢出
  简体   繁体   English

不同版本的R，lme4和OS X在glmer中给出不同的固定效应显着性结果

[英]Different versions of R, lme4 and OS X give different fixed-effects significance results in glmer

 原文  2014-09-06 11:59:40  9   1    r/ macos/ lme4/ mixed-models

问题描述

I am running a logit mixed-effects model using glmer() in package lme4. 我在包lme4中使用glmer（）运行logit混合效果模型。 The experiment used a within-subjects within-items design with Subjects and Items as crossed random effects. 该实验使用受试者内部项目设计，其中主题和项目为交叉随机效应。

My problem: different versions of R and lme4 (run on different OS X) produce different standard errors estimates for the fixed effects, and consequently, different significance results. 我的问题：不同版本的R和lme4（在不同的OS X上运行）对固定效果产生不同的标准误差估计，因此产生不同的显着性结果。

Here is a subset of my data (data from the last two subjects): 这是我的数据的子集（来自最后两个主题的数据）： 

structure(list(SubjN = c(87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 
87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 87L, 
87L, 87L, 87L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 
88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 88L, 
88L), Items = structure(c(3L, 10L, 11L, 5L, 1L, 12L, 2L, 6L, 
9L, 6L, 3L, 4L, 8L, 11L, 12L, 7L, 8L, 2L, 7L, 10L, 9L, 5L, 1L, 
4L, 10L, 3L, 5L, 11L, 12L, 1L, 2L, 6L, 9L, 6L, 3L, 4L, 8L, 11L, 
12L, 7L, 2L, 8L, 10L, 7L, 9L, 5L, 1L, 4L), .Label = c("a", "c", 
"k", "f", "g", "i", "d", "l", "e", "j", "b", "h"), class = "factor"), 
IV1 = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L), .Label = c("N", "L", "P"
), class = "factor"), DV = c(0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 
0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0), 
IV1.h = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L), contrasts = structure(c(-1, 
0.5, 0.5, 0, -0.5, 0.5), .Dim = c(3L, 2L), .Dimnames = list(
    c("N", "L", "P"), c("N_vs_L&P", "L_vs_P"))), .Label = c("N", 
"L", "P"), class = "factor"), N_vs_LP = c(-1, -1, -1, -1, 
-1, -1, -1, -1, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 
0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, -1, -1, -1, -1, -1, -1, 
-1, -1, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 
0.5, 0.5, 0.5, 0.5, 0.5, 0.5), L_vs_P = c(0, 0, 0, 0, 0, 
0, 0, 0, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, 
0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0, 0, 0, 0, 0, 0, 
0, 0, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, 0.5, 
0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5)), .Names = c("SubjN", 
"Items", "IV1", "DV", "IV1.h", "N_vs_LP", "L_vs_P"), row.names = c("3099", 
"3100", "3101", "3102", "3103", "3104", "3119", "3120", "3107", 
"3108", "3109", "3110", "3097", "3098", "3105", "3106", "3115", 
"3116", "3117", "3118", "3111", "3112", "3113", "3114", "3147", 
"3148", "3149", "3150", "3151", "3152", "3167", "3168", "3155", 
"3156", "3157", "3158", "3145", "3146", "3153", "3154", "3163", 
"3164", "3165", "3166", "3159", "3160", "3161", "3162"), class = "data.frame")

Each subject was tested on 24 trials on 3 different conditions (factor IV1, levels: N, L, P). 每个受试者在3个不同条件下进行24次试验（因子IV1，水平：N，L，P）。 I recorded whether they produced a target linguistic structure (DV == 1) or not (DV == 0). 我记录他们是否产生了目标语言结构（DV == 1）或不产生（DV == 0）。 In the analysis, I only included those subjects who produced the target structure at least one. 在分析中，我只包括那些产生目标结构的受试者至少一个。 Nonetheless, most of them produced the target structure only on very few occasion. 尽管如此，他们中的大多数只在极少数情况下产生了目标结构。 This is the proportion of DV == 1 produced by each subject in each condition: 这是每个条件下每个主题产生的DV == 1的比例： 

library(plyr)
#dput(ddply(mydata, .(SubjN, IV1), summarise, l = length(DV), y = round(mean(DV),2)))

structure(list(SubjN = c(1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 
4L, 4L, 4L, 5L, 5L, 5L, 6L, 6L, 6L, 7L, 7L, 7L, 8L, 8L, 8L, 9L, 
9L, 9L, 10L, 10L, 10L, 11L, 11L, 11L, 12L, 12L, 12L, 13L, 13L, 
13L, 14L, 14L, 14L, 15L, 15L, 15L, 16L, 16L, 16L, 17L, 17L, 17L, 
18L, 18L, 18L, 19L, 19L, 19L, 20L, 20L, 20L, 21L, 21L, 21L, 22L, 
22L, 22L, 23L, 23L, 23L, 24L, 24L, 24L, 25L, 25L, 25L, 26L, 26L, 
26L, 27L, 27L, 27L, 28L, 28L, 28L, 29L, 29L, 29L, 30L, 30L, 30L, 
31L, 31L, 31L, 32L, 32L, 32L, 33L, 33L, 33L, 34L, 34L, 34L, 35L, 
35L, 35L, 36L, 36L, 36L, 37L, 37L, 37L, 38L, 38L, 38L, 39L, 39L, 
39L, 40L, 40L, 40L, 41L, 41L, 41L, 42L, 42L, 42L, 43L, 43L, 43L, 
44L, 44L, 44L, 45L, 45L, 45L, 46L, 46L, 46L, 47L, 47L, 47L, 48L, 
48L, 48L, 49L, 49L, 49L, 50L, 50L, 50L, 51L, 51L, 51L, 52L, 52L, 
52L, 53L, 53L, 53L, 54L, 54L, 54L, 55L, 55L, 55L, 56L, 56L, 56L, 
57L, 57L, 57L, 58L, 58L, 58L, 59L, 59L, 59L, 60L, 60L, 60L, 61L, 
61L, 61L, 62L, 62L, 62L, 63L, 63L, 63L, 64L, 64L, 64L, 65L, 65L, 
65L, 66L, 66L, 66L, 67L, 67L, 67L, 68L, 68L, 68L, 69L, 69L, 69L, 
70L, 70L, 70L, 71L, 71L, 71L, 72L, 72L, 72L, 73L, 73L, 73L, 74L, 
74L, 74L, 75L, 75L, 75L, 76L, 76L, 76L, 77L, 77L, 77L, 78L, 78L, 
78L, 79L, 79L, 79L, 80L, 80L, 80L, 81L, 81L, 81L, 82L, 82L, 82L, 
83L, 83L, 83L, 84L, 84L, 84L, 85L, 85L, 85L, 86L, 86L, 86L, 87L, 
87L, 87L, 88L, 88L, 88L), IV1 = structure(c(1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L,      
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 
1L, 2L, 3L), .Label = c("N", "L", "P"), class = "factor"), l = c(8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 7L, 8L, 7L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 7L, 8L, 8L, 8L, 8L, 8L, 8L, 
7L, 8L, 6L, 7L, 7L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 7L, 8L, 8L, 7L, 7L, 8L, 7L, 8L, 
8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 6L, 8L, 4L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 7L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 8L, 7L, 
8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 
8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 8L, 7L, 8L, 7L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 8L), y = c(1, 0.88, 1, 0.5, 0.25, 0.62, 
0, 0, 0.25, 0, 0.25, 0, 0.12, 0, 0, 0, 0.12, 0, 0, 0.12, 0.12, 
0, 0, 0.12, 0.38, 0, 0.25, 0, 0.12, 0, 0.12, 0, 0.25, 0, 0, 0.12, 
0.5, 0.25, 0.5, 0, 0, 0.12, 0, 0.25, 0.12, 0, 0, 0.12, 0, 0.12, 
0, 0, 0.12, 0.12, 0.12, 0.62, 0, 0, 0.5, 0.25, 1, 0.88, 1, 0, 
0, 0.12, 0, 0.12, 0.12, 0.12, 0.12, 0, 0.62, 0.62, 0.38, 0.5, 
0.88, 0.12, 0.12, 0, 0, 0.12, 0.12, 0, 0, 0.12, 0, 0, 0.12, 0, 
0, 0.12, 0, 0, 0.25, 0, 0, 0.14, 0, 0.5, 0.57, 0.29, 0, 0.12, 
0, 0, 0.12, 0, 0.25, 0.5, 0.25, 0, 0.12, 0.12, 0.25, 0, 0.38, 
0, 0, 0.12, 0, 0, 1, 0.25, 0.12, 0.25, 0, 0.12, 0.12, 0, 0, 0.12, 
0, 0, 0.12, 0.12, 0, 0, 0.12, 0, 0.14, 0.14, 0.12, 0, 0.12, 0, 
0, 0.12, 0.12, 0, 1, 0.88, 1, 0, 0.12, 0, 0.12, 0, 0, 0.12, 0, 
0.12, 0, 0, 0.12, 0.12, 0.12, 0.12, 1, 1, 1, 0.12, 0, 0, 0.12, 
0.38, 0, 0, 0.12, 0, 0, 0, 0.5, 0.5, 0, 0.25, 0, 0.12, 0.29, 
0, 0, 0.38, 0, 0, 0.62, 0.5, 0, 0.12, 0, 0.12, 0.12, 0.25, 0.12, 
0.25, 0.12, 0, 0.12, 0, 0, 0.12, 0, 0, 0.12, 0, 0.12, 0.12, 0, 
0.12, 0.12, 0, 0, 0.12, 0.12, 0.12, 0, 0.38, 0.12, 0.57, 0, 0.12, 
0, 0, 0.12, 0, 0, 0.12, 0, 0, 0.12, 0.14, 0.88, 0.88, 0.86, 0, 
0, 0.14, 0, 0.12, 0.14, 0, 0.12, 0, 0, 0, 0.12, 0, 0, 0.12, 0.38, 
0, 0, 0.5, 0.12, 0)), .Names = c("SubjN", "IV1", "l", "y"), row.names = c(NA, 
-264L), class = "data.frame")

I run the following model including IV1 as fixed effect with helmert-contrast coding; 我运行以下模型，包括IV1作为具有helmert对比度编码的固定效果; first contrast: N vs. L & P, second contrast: L vs. P. 第一对比：N vs. L＆P，第二对比：L vs. P. 

m1 <- glmer(DV ~ IV1.h + (1 + IV1.h|SubjN) +  (1|Items) + (0 + N_vs_LP|Items) + (0 + L_vs_P|Items), family ='binomial', mydata)

The model does not allow for the correlation between the by-Items random variables (I did this by creating separate slopes for the two contrasts), since when correlation was allowed they were perfectly correlated (which I interpreted as a sign of over-parametrization). 该模型不允许逐项随机变量之间的相关性（我通过为两个对比创建单独的斜率来实现这一点），因为当允许相关时它们是完全相关的（我将其解释为过度参数化的标志） 。

1) Results using os x 10.8.5 mountain lion R version 3.0.2 (2013-09-25) lme4_1.0-5 (the original analysis I run) 1）结果使用os x 10.8.5山狮R版3.0.2（2013-09-25）lme4_1.0-5（我运行的原始分析） 

Generalized linear mixed model fit by maximum likelihood ['glmerMod']
 Family: binomial ( logit )
Formula: DV ~ IV1.h + (1 + N_vs_LP + L_vs_P | SubjN) + (1 | Items) + (0 + N_vs_LP | Items)     + (0 + L_vs_P | Items) 
   Data: mydata 

      AIC       BIC    logLik  deviance 
1492.5408 1560.2050 -734.2704 1468.5408 

Random effects:
 Groups  Name          Variance  Std.Dev. Corr       
 SubjN    (Intercept)   2.3885505 1.54549             
          N_vs_LP       0.4394195 0.66289  -0.69      
          L_vs_P        1.9287559 1.38880   0.04  0.08
 Items    (Intercept)   0.0531518 0.23055
 Items.1  N_vs_LP       0.0001950 0.01396
 Items.2  L_vs_P        0.0003619 0.01902             

Number of obs: 2077, groups: SubjN, 88; Items, 12

Fixed effects:
                              Estimate Std. Error z value Pr(>|z|)    
(Intercept)                    -2.2998     0.1964 -11.710  < 2e-16 ***
IV1.hN_vs_L&P                   0.3704     0.1378   2.689  0.00717 ** 
IV1.hL_vs_P                     0.2060     0.2320   0.888  0.37459    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Correlation of Fixed Effects:
              (Intr) IV1.N_
IV1.hN_vs_L&P -0.388       
IV1.hL_vs_P    0.014  0.019

2) Results using: OS X 10.9.4 Mavericks R version 3.1.1 (2014-07-10) lme4_1.1-7 optimizer 'bobyqa' 2）结果使用：OS X 10.9.4 Mavericks R 3.1.1版（2014-07-10）lme4_1.1-7优化器'bobyqa' 

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation)     ['glmerMod']
 Family: binomial  ( logit )
Formula: DV ~ IV1.h + (1 + N_vs_LP + L_vs_P | SubjN) + (1 | Items) + (0 +  
    N_vs_LP | Items) + (0 + L_vs_P | Items)
   Data: mydata
Control: glmerControl(optimizer = "bobyqa")

     AIC      BIC   logLik deviance df.resid 
  1492.5   1560.2   -734.3   1468.5     2065 

Scaled residuals: 
    Min      1Q  Median      3Q     Max 
-2.4174 -0.3364 -0.2595 -0.1706  4.6028 

Random effects:
 Groups  Name        Variance Std.Dev. Corr       
 SubjN   (Intercept) 2.38791  1.5453              
         N_vs_LP     0.43935  0.6628   -0.69      
         L_vs_P      1.92629  1.3879    0.04  0.07
 Items   (Intercept) 0.05319  0.2306              
 Items.1 N_vs_LP     0.00000  0.0000              
 Items.2 L_vs_P      0.00000  0.0000              
Number of obs: 2077, groups:  SubjN, 88; Items, 12

Fixed effects:
              Estimate Std. Error z value Pr(>|z|)    
(Intercept)    -2.2998     0.2095 -10.975   <2e-16 ***
IV1.hN_vs_L&P   0.3703     0.1892   1.958   0.0503 .  
IV1.hL_vs_P     0.2063     0.2679   0.770   0.4413    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Correlation of Fixed Effects:
            (Intr) IV1.N_
IV1.hN__L&P -0.379       
IV1.hL_vs_P -0.001  0.003

I really don't know which outcome I should trust. 我真的不知道应该相信哪个结果。 Any help would be very much appreciated. 任何帮助将非常感谢。

Ps. PS。 Sorry if something is not clear - it's my first post :) 对不起，如果事情不明确 - 这是我的第一篇帖子:)

Thanks very much! 非常感谢！

1 个解决方案

解决方案1
 6 已采纳  2014-09-06 12:41:26

From lme4 's NEWS file , for version 1.1-4 来自lme4的NEWS文件 ，版本为1.1-4

Standard errors of fixed effects are now computed from the approximate Hessian by default (see the use.hessian argument in vcov.merMod); 现在默认情况下从近似Hessian计算固定效应的标准误差（参见vcov.merMod中的use.hessian参数）; this gives better (correct) answers when the estimates of the random- and fixed-effect parameters are correlated (Github #47) 当随机和固定效应参数的估计相关时，这给出了更好（正确）的答案（Github＃47）

The description of the problem is here 问题的描述在这里

You should be able to retrieve the old standard errors from the newer (1.1-7) model by sqrt(diag(vcov(fitted_model,use.hessian=FALSE))) , but the new version is more likely to be correct. 您应该能够通过sqrt(diag(vcov(fitted_model,use.hessian=FALSE)))从较新的（1.1-7）模型中检索旧的标准错误，但新版本更可能是正确的。

For more precise confidence intervals/p values, you can do a likelihood ratio test (use anova to compare nested models) and/or compute the profile confidence intervals with confint(fitted_model,which="beta_") . 对于更精确的置信区间/ p值​​，您可以进行似然比检验（使用anova比较嵌套模型）和/或使用confint(fitted_model,which="beta_")计算轮廓置信区间confint(fitted_model,which="beta_") 。

暂无
暂无

声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

相关问题 不同 R/lme4 版本的奇异拟合的不匹配结果 - Mismatching results for singular fit with different R/lme4 versions 混合效应逻辑回归：MASS和lme4的不同结果 - mixed-effects logistic regression: different results with MASS and lme4 nlme和lme4中的不同随机效应 - different random effects in nlme and lme4 excel 的等效项 如果 r 中的 lme4 glmer 出现错误 - An equivalent for the excel If error for lme4 glmer in r 在 R 中指定逻辑混合 model 中重复测量的随机效应：lme4::glmer - Specifying random effects for repeated measures in logistic mixed model in R: lme4::glmer 如何在nlme与lme4中指定不同的随机效果？ - How to specify different random effects in nlme vs. lme4? 如何遍历列以评估连续 lme4 混合模型中的不同固定效应并提取系数和 P 值？ - How to loop over columns to evaluate different fixed effects in consecutive lme4 mixed models and extract the coefficients and P values? R 中 function 的时间固定效应 - Time fixed-effects with felm function in R 不同的lme4版本给出不同的结果 - Different lme4 version is giving different result 如何从lme4中提取固定效果的一部分摘要？ - How to extract fixed effects part of summary from lme4?
相关标签
 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM