评估嵌套效应的成对比较

时间:2016-10-09 23:20:20

标签: r anova lsmeans

我有一个模型,我试图调查一些嵌套效果的成对比较。我不确定我是否正确地编写了模型,而且我不理解如何实际评估嵌套术语。

我的数据框有一个名为' quality'的响应变量。和三个预测变量称为" site" "一个月"和" day"。在我的实验设置中,我测量了每个人的质量。有两个网站。我在4个月内对每个网站进行了抽样每个月都有连续4天的抽样。我想知道一天中的个人是否与同一个月内其他日子的人的质量明显不同。我没有兴趣比较另一个月的天数到一个月到几天。

我的数据框如下

   test<- structure(list(Site = structure(c(2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("H", "W"), class = "factor"), 
Day = structure(c(19L, 19L, 19L, 19L, 19L, 19L, 19L, 19L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 5L, 5L, 
5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 
18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 15L, 15L, 15L, 
15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 
15L, 15L, 15L, 15L, 15L, 15L, 15L, 15L, 26L, 26L, 26L, 26L, 
26L, 26L, 26L, 26L, 26L, 26L, 26L, 26L, 26L, 26L, 26L, 26L, 
26L, 26L, 8L, 8L, 8L, 8L, 8L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 
17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 17L, 
17L, 17L, 17L, 17L, 14L, 14L, 14L, 14L, 14L, 14L, 14L, 14L, 
14L, 14L, 14L, 14L, 14L, 25L, 25L, 25L, 11L, 11L, 11L, 11L, 
11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 
11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 11L, 22L, 
22L, 7L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 
16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 16L, 
16L, 16L, 16L, 16L, 16L, 16L, 13L, 13L, 13L, 13L, 13L, 13L, 
13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 
13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 24L, 24L, 24L, 
24L, 24L, 24L, 24L, 24L, 24L, 24L, 24L, 24L, 24L, 24L, 10L, 
10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 
10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 10L, 
10L, 10L, 10L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 
21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 21L, 
21L, 21L, 21L, 21L, 21L, 21L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 
6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 12L, 
12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 
12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 
23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 
23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 23L, 
23L, 23L, 23L, 23L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 
9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 
9L, 9L, 9L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 
20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 20L, 
20L, 20L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L), .Label = c("H1", "H10", "H11", "H12", "H13", "H14", 
"H15", "H16", "H2", "H3", "H4", "H6", "H7", "H8", "H9", "W10", 
"W11", "W12", "W13", "W2", "W3", "W4", "W6", "W7", "W8", 
"W9"), class = "factor"), Month = structure(c(3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 2L, 2L, 2L, 2L, 2L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("August", 
"November", "October", "September"), class = "factor"), Quality = c(42.535, 
46.651, 45.466, 43.483, 44.896, 46.581, 47.494, 47.529, 46.562, 
45.111, 45.982, 48.367, 47.39, 45.388, 46.313, 44.732, 48.641, 
46.614, 45.234, 45.96, 44.795, 44.333, 46.559, 46.826, 44.166, 
45.19, 46.661, 45.481, 46.828, 43.487, 49.505, 48.558, 45.218, 
44.802, 43.975, 47.23, 44.85, 46.213, 44.726, 43.036, 47.211, 
45.536, 44.62, 44.297, 36.115, 39.314, 42.349, 44.919, 46.296, 
46.317, 45.858, 45.036, 45.861, 48.85, 45.337, 45.03, 47.4, 
48.78, 49.829, 45.12, 45.599, 43.235, 44.735, 44.889, 45.666, 
46.475, 44.888, 46.215, 42.242, 46.341, 45.992, 43.549, 46.612, 
44.232, 42.706, 42.064, 43.837, 43.351, 41.064, 44.364, 42.597, 
45.561, 44.51, 45.184, 44.896, 45.772, 47.43, 44.08, 44.697, 
45.141, 43.776, 47.175, 46.115, 43.39, 47.426, 47.636, 43.672, 
45.987, 45.338, 46.644, 42.192, 47.011, 45.856, 44.764, 36.285, 
33.741, 34.324, 35.101, 46.844, 42.52, 48.649, 44.364, 44.688, 
45.822, 44.945, 44.311, 44.684, 42.787, 45.516, 46.16, 46.289, 
45.661, 45.772, 43.845, 48.717, 46.567, 44.719, 46.585, 45.33, 
45.995, 48.053, 44.734, 51.233, 44.597, 45.742, 46.567, 46.478, 
44.382, 47.316, 46.205, 45.111, 47.575, 46.014, 44.533, 45.347, 
45.983, 47.053, 44.855, 48.021, 45.155, 49.248, 45.634, 48.815, 
45.413, 43.091, 47.854, 45.19, 47.495, 47.323, 48.076, 44.183, 
43.182, 46.267, 41.58, 44.237, 45.607, 48.517, 44.639, 44.773, 
42.787, 43.965, 46.629, 46.256, 47.688, 44.126, 44.712, 47.097, 
44.561, 47.306, 45.323, 46.328, 45.832, 46.075, 46.778, 47.445, 
45.582, 47.691, 45.193, 48.453, 46.301, 44.847, 43.675, 46.066, 
47.896, 45.2, 44.959, 47.401, 46.267, 45.743, 47.411, 46.926, 
46.24, 46.212, 44.988, 36.552, 38.027, 47.355, 40.147, 38.094, 
39.043, 37.589, 46.491, 46.413, 43.92, 45.228, 46.319, 44.764, 
47.376, 43.924, 45.203, 45.418, 45.684, 46.34, 43.655, 44.365, 
46.927, 48.269, 45.473, 46.451, 42.752, 48.346, 47.832, 46.534, 
46.47, 43.282, 47.749, 44.856, 46.551, 45.925, 45.669, 47.263, 
44.367, 47.017, 42.922, 44.904, 48.85, 45.535, 48.512, 46.154, 
47.306, 46.571, 46.619, 46.092, 43.808, 47.7, 48.482, 44.407, 
45.442, 44.771, 46.373, 47.777, 43.012, 46.154, 45.203, 46.443, 
43.461, 45.714, 40.776, 48.949, 45.72, 48.269, 45.782, 43.945, 
45.382, 43.729, 44.187, 45.267, 46.012, 42.234, 43.431, 41.973, 
45.597, 45.993, 46.303, 44.493, 44.981, 46.487, 45.01, 47.009, 
46.904, 48.277, 48.585, 48.625, 47.511, 44.011, 42.21, 47.124, 
44.244, 47.76, 47.299, 45.278, 45.564, 44.621, 46.75, 45.396, 
44.947, 46.185, 45.399, 46.095, 49.545, 47.211, 43.613, 48.494, 
44.102, 45.888, 45.473, 47.222, 46.681, 45.863, 47.834, 48.386, 
46.979, 46.318, 46.061, 46.347, 47.976, 47.079, 48.254, 47.643, 
46.244, 46.717, 44.574, 45.177, 44.879, 46.485, 47.416, 50.235, 
45.626, 48.117, 44.529, 44.281, 47.087, 47.356, 43.234, 45.841, 
43.487, 42.997, 35.322, 45.554, 44.973, 43.396, 43.023, 44.65, 
47.088, 41.934, 45.704, 44.559, 37.969, 42.687, 42.995, 45.287, 
45.21, 43.335, 46.892, 45.534, 44.19, 43.606, 44.173, 49.334, 
44.888, 47.477, 47.054, 41.041, 46.629, 45.049, 44.478, 40.278, 
43.044, 43.575, 46.194, 42.688, 41.361, 46.828, 45.534, 47.395, 
45.431, 45.433, 45.331, 43.947, 47.371, 48.308, 45.726, 41.833, 
45.782, 44.756, 45.406, 45.661, 43.447, 46.932, 45.495, 44.349, 
40.493, 43.603, 48.151, 44.037, 44.379, 45.934, 44.854, 42.321, 
46.198, 44.622, 46.077, 45.306, 48.951, 47.972, 42.581, 43.608, 
45.988, 44.955, 45.097, 46.768, 44.722, 45.971, 46.612, 48.956, 
47.669, 47.757, 47.189, 44.184, 48.464, 49.546, 48.021, 45.448, 
45.573, 46.778, 45.769, 45.419, 45.277, 47.489, 46.762, 46.238, 
47.509, 47.249, 46.243, 46.124, 46.801, 47.385, 43.614, 44.661, 
45.96, 48.791, 47.872, 42.402, 45.651, 45.927, 43.781, 49.923, 
47.153, 46.87, 43.767, 47.3, 46.897, 44.932, 45.135, 50.124, 
45.366, 45.063, 45.958, 46.731, 43.863, 45.095, 47.755, 45.446, 
45.145, 45.998, 46.377, 44.369, 46.485, 48.852, 45.365, 45.934, 
44.856, 48.195, 45.424, 49.05, 46.115, 43.077, 48.305, 44.784, 
44.934, 46.253, 46.203, 48.36, 47.36, 48.872, 44.803)), .Names = c("Site", 
"Day", "Month", "Quality"), class = "data.frame", row.names = c(NA, 
-496L)) 

我已经写了这样的模型

    library(lsmeans)
fit1<-aov(Quality~Site*Month + (Site*Month)/Day, data=test)

这个模型似乎对我有用。我理解如何评估网站和月份的交互术语和主要影响,但我出于某种原因努力评估日期。我试过了

dayeffects<-lsmeans(fit1, pairwise~Site*Month/Day, adjust="bonferroni")
results <- dayeffects[[2]]
summary(results)[!is.na(summary(results)[,4]),] 

但这似乎测试了每个成对比较,而不是遵循嵌套结构。就像我上面说的那样,我只想比较同一个月和网站内发生的日子。

虽然我知道我可以从上面进行比较,但我觉得我可能做错了。它也使得bonferroni调整过度。

任何帮助都会很棒。感谢

1 个答案:

答案 0 :(得分:0)

好的,我清除了我的想法。您可以在SiteMonth之间按lsmeans(model, pairwise ~ Day | Site * Month )对它们进行比较。 (我使用~ (Site*Month)/Day作为模型,但在此主题中,~ Site + Month + Site:Month:Day~ Site * Month * Day会返回相同的结果,因为您的Day是唯一的。)

fit <-aov(Quality ~ (Site*Month)/Day, data=test)      # this model is equivalent to OP's one.
res <- lsmeans(fit, pairwise ~ Day | Site * Month, adjust="bonferroni")
results <- summary(res[[2]])[!is.na(summary(res[[2]])[,4]),]

> results[25:30,]
Site = H, Month = September:
 contrast   estimate        SE  df t.ratio p.value
 H6 - H7  -1.0626904 0.5348000 470  -1.987  0.2850
 H6 - H8  -0.1373969 0.6588578 470  -0.209  1.0000
 H6 - H9   0.3862017 0.5567090 470   0.694  1.0000
 H7 - H8   0.9252934 0.6504561 470   1.423  0.9332
 H7 - H9   1.4488921 0.5467399 470   2.650  0.0499
 H8 - H9   0.5235987 0.6685859 470   0.783  1.0000
检查p.adjustment
res0 <- lsmeans(fit, pairwise ~ Day | Site * Month, adjust="none")
results0 <- summary(res0[[2]])[!is.na(summary(res0[[2]])[,4]),]  # get non-adjusted p.value
res0.p <- p.adjust(results0$p.value[25:30], "bonferroni")    # semi-manually p.adjustment
identical(results$p.value[25:30], res0.p)                    # [1] TRUE