我有一个大型数据集，其中包含数千个长度和重量的测量值。我在这里提供了500个观察结果的子集：
df <- structure(list(length_cm = c(24.7, 23.8, 21.9, 23.2, 23.5, 22.2,
20.5, 22.6, 24, 21.6, 22.4, 21.2, 20.6, 23.1, 21.4, 23.1, 23.5,
23, 21.8, 22.4, 23, 23.8, 24, 21, 23.4, 23.2, 21.6, 25.9, 22.1,
30.6, 22.1, 21.7, 23.2, 21.1, 23.8, 23.2, 27.2, 23.8, 21.6, 21.1,
21.7, 22.9, 23.3, 24.1, 22.7, 20.4, 22.5, 21.7, 23.2, 22.7, 20.6,
23.7, 24.6, 23.5, 26.3, 23.6, 22.2, 23.6, 21.4, 23.3, 24.7, 24.4,
21.8, 24.9, 22.2, 23.1, 25, 23.5, 22.5, 20.4, 23.9, 23.7, 24,
24.2, 22.9, 36.4, 30, 26, 28.5, 27, 35.7, 24.3, 28.6, 29.8, 18.7,
25.7, 34.7, 31.4, 23.4, 37.7, 26.7, 28.3, 30.8, 29.2, 27.2, 25.6,
39, 35.1, 41.2, 35.7, 29.9, 25.7, 24.6, 24, 24.9, 31, 29.9, 29.4,
25.4, 20.2, 27.8, 32.7, 23.4, 29.1, 26.3, 25.7, 26, 24.9, 26.3,
31.5, 30.1, 25.9, 28.8, 37.9, 38.4, 21.5, 20.5, 21.3, 21.3, 20.9,
20.8, 22.5, 22.4, 21.4, 16.8, 17.3, 22.7, 19.7, 21.2, 18.1, 23.5,
18.1, 22, 18.5, 18.4, 19.2, 19.4, 19.9, 20.5, 18.6, 22.6, 20.9,
20.7, 20.6, 20.6, 21.6, 23.7, 22.8, 22.9, 20.8, 21.3, 23.5, 21.1,
21.6, 24, 21, 23.3, 20.3, 22.4, 23.7, 24.6, 20.7, 23.1, 22.6,
22.7, 19.5, 23, 19.8, 21, 19.8, 19.8, 17.2, 21.8, 25.3, 21.3,
19.2, 22.1, 24.5, 23.2, 22.6, 19, 22, 17.5, 19.9, 24.4, 23.7,
19.9, 23, 20.5, 18.3, 23.2, 21.1, 20.4, 22.2, 19.7, 19.2, 24,
23.3, 23.3, 19, 21.5, 22, 19.1, 23.7, 19.9, 21.2, 23, 27.3, 20.7,
22, 19.3, 24.9, 18.2, 20, 19.3, 25, 18, 21.8, 23.4, 23.9, 25.2,
18.5, 22.2, 24.6, 22, 20.4, 20.7, 21.7, 19.1, 23.1, 21.5, 21.2,
20.6, 22.3, 22.8, 21.3, 21.6, 22, 23, 24.2, 21.3, 19.7, 18.8,
20.9, 20.3, 22.3, 18.9, 19.9, 20.2, 23.9, 19.7, 19.5, 17.6, 23.1,
20.4, 20, 19.7, 20.3, 21.2, 23.9, 24, 25.6, 23.9, 23.5, 20.5,
30.8, 32.8, 28.4, 28.7, 28, 28.9, 29.8, 31, 31.7, 28.6, 28.7,
28.7, 26.7, 24.6, 30, 36.5, 26.5, 32, 29.6, 30.7, 27.7, 24.1,
29.8, 28.8, 26, 22.4, 24, 24.8, 22.7, 22.7, 23.8, 25.3, 32.3,
26.8, 22.1, 24.2, 23.8, 25.3, 24.1, 22.6, 22.9, 24.4, 26.7, 24.4,
24.7, 25, 23.7, 24.3, 22.3, 22.7, 20, 22.5, 24.5, 25.1, 24, 22,
20, 21.9, 18.3, 19.9, 19.4, 23.5, 20.2, 20, 17.8, 20.5, 23.2,
18.5, 21.2, 18.2, 19.1, 22.1, 18.3, 21.6, 19.5, 22.7, 23.6, 24.6,
23.2, 24.4, 19.1, 22.8, 23, 18.8, 22.6, 19, 21.7, 20.8, 23.7,
20.8, 20, 23.2, 22, 21.4, 20.6, 22.6, 23.8, 21, 26.4, 24.5, 32.6,
36.1, 36, 31, 33.1, 31.3, 34.2, 41.9, 35.4, 33.9, 31.9, 29.3,
34.2, 29.9, 36.4, 38.5, 30.7, 40.2, 34.1, 29.7, 37.8, 37.8, 35.3,
39, 39.5, 34.1, 30.5, 33.3, 33.2, 36, 31.6, 35, 34.2, 33.1, 31.5,
33.5, 33.7, 39, 33.2, 35, 34.1, 32.6, 36.2, 34.4, 31.7, 32, 37.5,
31.5, 32.7, 31.7, 35.7, 32.4, 28.5, 33.7, 33.9, 33.6, 34, 32,
29.8, 35, 36, 31.7, 32.5, 32, 31, 29.5, 33.4, 32.5, 26.5, 28,
35.3, 26, 26.5, 38.9, 32.7, 36.4, 35.7, 27.7, 25.8, 25.3, 30.1,
36, 33.4, 37, 33.6, 31.7, 29.7, 35.9, 28.5, 33.1, 33.9, 29, 36.5,
35.5, 29.2, 37.3, 40.3, 35.7, 32.6, 38.8, 40, 38.9, 39, 33.3,
33.5, 34.3, 38.8, 34.4, 36, 35.9, 35.1, 30.7, 38.1, 31.3, 35,
36.3, 32.4, 32.3, 35.5, 36.4, 36, 40.8, 34.2, 30.1, 35.6), wt_kg = c(0.165,
0.1412, 0.1043, 0.1225, 0.1247, 0.1099, 0.087, 0.1176, 0.1431,
0.1041, 0.1213, 0.0937, 0.0856, 0.1255, 0.1099, 0.124, 0.1361,
0.1384, 0.1021, 0.1113, 0.12, 0.1513, 0.1448, 0.0978, 0.138,
0.1232, 0.0942, 0.1881, 0.1038, 0.3498, 0.1122, 0.094, 0.1268,
0.1009, 0.1358, 0.12, 0.2388, 0.1456, 0.0982, 0.0903, 0.1005,
0.1252, 0.1138, 0.1476, 0.1326, 0.0849, 0.108, 0.0996, 0.1229,
0.1279, 0.0874, 0.1492, 0.1416, 0.1187, 0.193, 0.1383, 0.1125,
0.1449, 0.0941, 0.1265, 0.1823, 0.1455, 0.0948, 0.1603, 0.1119,
0.1124, 0.1641, 0.1259, 0.116, 0.086, 0.1361, 0.1284, 0.1403,
0.1461, 0.1195, 0.5985, 0.3099, 0.1829, 0.2688, 0.2244, 0.6214,
0.1554, 0.2475, 0.2976, 0.0683, 0.1731, 0.4751, 0.356, 0.1388,
0.5939, 0.2122, 0.2784, 0.3689, 0.3127, 0.2284, 0.1775, 0.6697,
0.5998, 0.8374, 0.5647, 0.3187, 0.1704, 0.1619, 0.1413, 0.1621,
0.3577, 0.319, 0.2846, 0.1815, 0.0776, 0.2567, 0.4483, 0.1337,
0.2798, 0.202, 0.1847, 0.1758, 0.1659, 0.1828, 0.3669, 0.3211,
0.1863, 0.2559, 0.6901, 0.6483, 0.0922, 0.088, 0.099, 0.0836,
0.094, 0.099, 0.1157, 0.1138, 0.1046, 0.0495, 0.0513, 0.119,
0.0761, 0.0936, 0.0564, 0.1438, 0.0636, 0.1134, 0.0641, 0.0594,
0.0713, 0.0733, 0.0804, 0.0853, 0.0689, 0.118, 0.0892, 0.0875,
0.0837, 0.0807, 0.1065, 0.1385, 0.1163, 0.1305, 0.0923, 0.0974,
0.1176, 0.0848, 0.1059, 0.157, 0.0932, 0.1127, 0.0779, 0.1048,
0.1327, 0.1688, 0.1096, 0.1304, 0.1173, 0.115, 0.0742, 0.129,
0.0629, 0.0992, 0.0758, 0.0722, 0.0535, 0.0958, 0.1721, 0.1017,
0.0766, 0.1099, 0.152, 0.128, 0.1185, 0.065, 0.1176, 0.0565,
0.0866, 0.163, 0.12, 0.0825, 0.1149, 0.0839, 0.0587, 0.1335,
0.0968, 0.0901, 0.1073, 0.0802, 0.0744, 0.1493, 0.1384, 0.1128,
0.0738, 0.1146, 0.1108, 0.08, 0.1285, 0.0829, 0.1116, 0.1368,
0.2348, 0.0995, 0.0989, 0.0748, 0.1484, 0.0629, 0.0823, 0.075,
0.1768, 0.0607, 0.1142, 0.1289, 0.1506, 0.1742, 0.0626, 0.1187,
0.1509, 0.1144, 0.0928, 0.0946, 0.099, 0.0717, 0.1318, 0.1025,
0.093, 0.0972, 0.1325, 0.1209, 0.0943, 0.1006, 0.1073, 0.1336,
0.1439, 0.1066, 0.0765, 0.0673, 0.1082, 0.0923, 0.1139, 0.068,
0.0758, 0.0868, 0.1499, 0.0779, 0.0794, 0.0575, 0.1392, 0.0915,
0.0845, 0.086, 0.084, 0.1049, 0.1486, 0.1573, 0.177, 0.1319,
0.13, 0.0872, 0.388, 0.4751, 0.2898, 0.2931, 0.2663, 0.2838,
0.3494, 0.3675, 0.4342, 0.2907, 0.3072, 0.2815, 0.2761, 0.1945,
0.3512, 0.615, 0.2195, 0.4818, 0.3684, 0.4056, 0.2841, 0.1617,
0.3425, 0.288, 0.1962, 0.1285, 0.1553, 0.1708, 0.1332, 0.1167,
0.1491, 0.2028, 0.1267, 0.2406, 0.1257, 0.1499, 0.1559, 0.1895,
0.1508, 0.1111, 0.1274, 0.1675, 0.2324, 0.1732, 0.1491, 0.1568,
0.1465, 0.1548, 0.1245, 0.1399, 0.0855, 0.1151, 0.1612, 0.1693,
0.1493, 0.1208, 0.088, 0.1106, 0.0654, 0.0827, 0.0794, 0.1331,
0.0834, 0.0837, 0.0619, 0.092, 0.1397, 0.071, 0.1035, 0.0676,
0.0729, 0.0906, 0.064, 0.0985, 0.0823, 0.1206, 0.155, 0.1438,
0.1357, 0.1695, 0.0834, 0.1359, 0.1289, 0.0764, 0.1249, 0.0775,
0.1139, 0.104, 0.1566, 0.1069, 0.0869, 0.1376, 0.1223, 0.105,
0.0996, 0.1356, 0.1335, 0.0951, 0.2162, 0.1744, 0.4547, 0.5789,
0.5555, 0.3899, 0.5037, 0.4281, 0.486, 1.0209, 0.5855, 0.5312,
0.488, 0.3133, 0.5054, 0.3724, 0.59, 0.8119, 0.3811, 0.797, 0.5139,
0.348, 0.7722, 0.743, 0.548, 0.8791, 0.9054, 0.5392, 0.4333,
0.5314, 0.4976, 0.5953, 0.4288, 0.5179, 0.5634, 0.5331, 0.4371,
0.5709, 0.5065, 0.8047, 0.5368, 0.5657, 0.5816, 0.4763, 0.5907,
0.533, 0.4384, 0.4949, 0.7277, 0.4445, 0.4894, 0.4655, 0.5384,
0.5106, 0.3343, 0.5186, 0.5262, 0.5311, 0.495, 0.4691, 0.3465,
0.5558, 0.5975, 0.4768, 0.4802, 0.4573, 0.4037, 0.3316, 0.5152,
0.4673, 0.2356, 0.2905, 0.5672, 0.2097, 0.2216, 0.7384, 0.4089,
0.6159, 0.5219, 0.2866, 0.2443, 0.2071, 0.3658, 0.5861, 0.5021,
0.6953, 0.5053, 0.3978, 0.3853, 0.6207, 0.2944, 0.507, 0.4412,
0.3424, 0.6597, 0.5892, 0.3295, 0.6505, 0.9334, 0.6674, 0.4919,
0.8392, 0.9123, 0.813, 0.8223, 0.5801, 0.5745, 0.5148, 0.8514,
0.5563, 0.6417, 0.6445, 0.5701, 0.4186, 0.8303, 0.46, 0.6041,
0.6537, 0.5221, 0.4782, 0.5657, 0.6499, 0.6667, 0.9074, 0.555,
0.6696, 0.6083)), .Names = c("length_cm", "wt_kg"), row.names = c(NA,
500L), class = "data.frame")
长度和重量之间的关系不是线性的。不幸的是，我无法在此处包括整个数据集，但是当使用整个数据集时，gam提供了最佳的拟合度，这与建议黄土的子集不同。
我想专注于gam，因为我所追求的是适用于整个数据集的答案。
显然，即使在提供的子集中，我的数据也有一些异常值，在示例数据集（df）中，至少有两个明显的异常值。
library(ggplot2)
ggplot(df, aes(x=wt_kg, y=length_cm))+
geom_point()+
stat_smooth(method = "gam", formula = y ~ s(x), size = 1)
采用gam方法前进，我想生成预测间隔，以便我可以确定哪些点落入和偏离了95％的预测间隔。 这对于使用预测的线性回归非常简单：
l_model <- lm(wt_kg ~ length_cm, data=df)
df <- cbind(df, predict(l_model, interval = "prediction"))
然后简单地绘制间隔的上限和下限
ggplot(df, aes(y=wt_kg, x=length_cm)) +
geom_ribbon(aes(ymin = lwr, ymax = upr),
fill = "blue", alpha = 0.2) +
geom_point()
但是我似乎找不到使用gam而不是lm的类似方法。我已经尝试过mgcv软件包中的predict.gam，但没有成功。
library(mgcv)
df_model <- gam(wt_kg ~ length_cm, data=df)
gam_pred <- cbind(df, mgcv::predict.gam(df_model))
运行此程序时我没有收到任何错误，但是我得到的只是单个数据列，我不确定该如何解释。任何帮助将不胜感激。
答案 0 :(得分：0)
我认为您的代码的一部分是：
require(broom)
require(gam)
mod <- gam(wt_kg ~ length_cm, data=df)
pred <- augment(mod)
但是我不明白第二个ggplot2。 “ Pred” 具有拟合值和有关回归的其他功能，主要是 .resid