基于样本占比与均值的数据变量过滤技术问询

以下是针对你的数据集完成两步变量过滤的R代码：

# 加载你的数据集
df <- structure(list(Samples = 1:29, c01 = c(20.54997, 0, 0, 0, 0.00226, 
0, 0, 0, 0, 65.24078, 0, 0, 0, 0, 0, 0, 0, 12.83106, 0.38555, 
16.45834, 0, 0, 0.00064, 0.005, 0, 0, 0, 12.36956, 0), c02 = c(14.43141, 
0.25318, 0.83121, 0.54282, 8.50687, 0.63261, 1.16257, 0, 0.25563, 
1.55071, 0, 0.65672, 0.44498, 2.13339, 3.01204, 0.25696, 0.10432, 
0.0051, 0.00139, 0.11766, 0.57905, 0, 0.12239, 0, 0, 0, 1.82955, 
0.06883, 0), c03 = c(12.45032, 7.88459, 0.99032, 1.84241, 0.93828, 
2.26896, 3.33106, 8.50696, 5.13512, 1.582, 0.24858, 0.7017, 1.2043, 
2.58593, 1.02544, 0.75702, 1.96771, 5.16734, 3.82884, 1.55069, 
0.97391, 2.47482, 1.06835, 0.01934, 4.41235, 3.68062, 7.01643, 
1.24734, 7.71274), c04 = c(7.5127, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
0, 0.44856, 3.05223, 7.78115, 0, 0, 6.65269, 0, 0, 4.11219, 0, 
0, 1.00195, 0.00198, 0, 0, 11.43261, 0.00137, 0), c05 = c(6.57903, 
1.45815, 0.28668, 1.66816, 1.66008, 0.77155, 1.00269, 1.03546, 
0.60667, 1.46041, 1.35261, 0.0228, 1.38451, 1.51657, 4.8712, 
5.31965, 1.85348, 1.80051, 0.91537, 3.12064, 3.00647, 0.55887, 
0.02098, 5.04716, 0.66621, 0, 0.10575, 0.89721, 0.04418), c06 = c(3.29649, 
2.45599, 35.45449, 6.43284, 10.45324, 2.58769, 1.32091, 1.75716, 
0.11397, 2.06348, 8.57203, 12.7673, 8.50063, 10.19027, 12.73432, 
3.195, 8.80226, 6.47543, 2.56102, 2.7203, 4.73366, 2.73383, 0, 
0.25893, 4.73464, 0.00275, 5.13602, 18.49032, 0.91792), c07 = c(3.13373, 
0.13374, 1.15662, 0.00969, 0, 0.95451, 0.35991, 0.80672, 0, 0, 
2.07513, 0, 1.14085, 0, 0.19192, 0.16836, 0, 0.03299, 0.94231, 
0.49989, 0.32173, 0, 0, 0, 0.01308, 0.41972, 0.00371, 0, 0.19803
), c08 = c(2.42942, 0, 0.05228, 0, 0, 0.12539, 0, 0, 0, 0, 0, 
0, 0, 0, 0, 0.14092, 0, 0, 0, 0, 0, 0.06318, 0, 0, 0, 0, 0, 0, 
0), c09 = c(2.08329, 1.48825, 0.3651, 0.04518, 8.95546, 0.50419, 
2.08877, 1.59026, 6.83207, 0.01179, 0.03375, 2.25788, 5.37218, 
1.57857, 1.10407, 0.33614, 1.17234, 0.03309, 0.44883, 1.99542, 
4.34365, 3.09612, 0.64867, 0.00276, 3.6342, 13.58322, 1.21404, 
0.5201, 2.25059), c10 = c(2.01871, 0.16123, 0.17942, 0.01882, 
0, 0.07479, 0.12062, 0.04417, 0.06655, 0.44124, 0.94579, 0.42188, 
0.16252, 0, 0.07415, 0.37692, 0.05604, 0.02513, 0.01643, 0.36103, 
0.11106, 0.00976, 13.86118, 1.00675, 0.08545, 0.05809, 0.00459, 
0.04065, 3.15904), c11 = c(1.80767, 2.60171, 1.52571, 1.42421, 
0.54515, 3.10199, 6.42356, 0.70189, 1.63704, 1.37722, 0.34205, 
1.74508, 1.37933, 0.33491, 5.51428, 0.38707, 0, 3.56014, 3.03788, 
1.36028, 3.95165, 0.45644, 0.001, 0, 0.79523, 0.79666, 0.07401, 
1.9843, 0.1594), c12 = c(1.5703, 0.01536, 0.28295, 0.15292, 0.01971, 
1.43474, 0, 0, 0.0153, 0, 0.0923, 0.43918, 0.09941, 0, 0.20089, 
0.04423, 0, 0.92555, 0.22799, 0.46793, 0.54228, 0.99488, 0, 0, 
0.5476, 0, 0, 0.90751, 0), c13 = c(1.5414, 0.10995, 0.07654, 
0.21751, 0.95283, 0.42834, 0.00731, 0.01768, 0, 0.0078, 0.23865, 
0.19046, 0.3389, 0, 0.09742, 0.01043, 0.04515, 0.04054, 0.01455, 
0.08685, 0.02306, 0, 0, 0, 0.08105, 0, 0.18324, 0.0479, 0.01374
), c14 = c(1.44565, 2.00101, 0.06163, 0.94796, 0.53368, 2.67534, 
0.70291, 0.40036, 1.91345, 2.12895, 1.12242, 1.94325, 3.10206, 
1.99831, 1.43953, 0.12424, 0.49666, 0.31405, 0.10889, 2.56575, 
1.76696, 3.09154, 0, 6.71776, 1.80677, 0.31812, 0.04629, 1.62853, 
0.53896), c15 = c(1.42826, 0.4252, 0.00057, 0.10527, 0, 0, 1.34473, 
0.28319, 0.06461, 0, 0, 0, 0, 0, 0.60115, 0, 0, 2.88822, 0.45054, 
0, 0.02099, 0, 0, 0, 0, 0, 0, 1.36053, 0), c16 = c(1.42179, 0, 
0, 0, 0, 0, 0, 0, 0, 0.04243, 0, 0, 0, 0, 0, 0.53077, 0.37179, 
0, 0.16018, 1.09049, 0, 0, 0, 0, 0.5307, 0, 0, 0, 1.92085), c17 = c(1.24998, 
0.62542, 0.79644, 0.07306, 0.05504, 0.16539, 0.42176, 0.21835, 
0.03788, 0.20085, 0.27482, 0, 0.5395, 0, 0.81909, 1.48779, 1.01917, 
0.17791, 0.3074, 0.66899, 0.31369, 0.09243, 0, 2.60673, 0.22892, 
0, 0.04191, 0.44308, 2.61117)), class = "data.frame", row.names = c(NA, 
-29L))

# 第一步：保留在超过5%样本中取值>0的变量
# 计算每列（排除Samples）中>0的样本占比
non_zero_proportion <- colMeans(df[, -1] > 0)
# 筛选占比>5%的变量
selected_step1 <- names(non_zero_proportion[non_zero_proportion > 0.05])
# 构建第一步过滤后的数据集
df_step1 <- df[, c("Samples", selected_step1)]

# 第二步：剔除均值>0.2的变量
# 计算第一步后各变量的均值
col_averages <- colMeans(df_step1[, -1])
# 筛选均值≤0.2的变量
selected_step2 <- names(col_averages[col_averages <= 0.2])
# 最终过滤后的数据集
df_final <- df_step1[, c("Samples", selected_step2)]

# 查看最终保留的变量
names(df_final)

步骤说明 #

1. 第一步过滤：

   • 排除Samples列，用colMeans(df[, -1] > 0)计算每个变量中取值大于0的样本占比（逻辑值TRUE被视为1，FALSE视为0，均值即占比）。
   • 筛选出占比超过5%的变量，将这些变量与Samples列组合得到第一步结果。

2. 第二步过滤：

   • 基于第一步的结果，再次排除Samples列，计算剩余变量的均值。
   • 剔除均值大于0.2的变量，保留均值≤0.2的变量，得到最终的过滤数据集。

内容的提问来源于stack exchange，提问作者user14845003