Search code examples
rt-testhypothesis-test

Multiple t-test on independent group with a large dataframe


I've seen many similar posts but the vast majority of them are at least 3 years old and I'm not really sure they apply to my situations, so here we go.

A colleague asked for my help on a multiple t-test on her project.

Basically she has 20 observation x 30 variable dataframe that looks like this: | Group | Lipid 1 | Lipid 2 | ... | Lipid 28|

| -------- | -------------- |

| A | |B | | | |B |

What we want to do is a group comparison of each lipide (meaning a t-test for Lipide 1 between group A and B, then a t-test for Lipide 2 and so on).

We do not want to compare Lipids between them.

And of course, we'd like to not have to copy/paste the same 3 lines of code, especially since we've got 2 other dataframe with the same variable but different conditions.

I've tried one solution I saw in here but it gives me an error I'm not sure to understand:

sapply(foetal[,2:20], function(i) t.test(i ~ foetal$ID)) 
Error in if (stderr < 10 * .Machine$double.eps * max(abs(mx), abs(my))) stop("data are essentially constant") :  missing value where TRUE/FALSE needed In addition: Warning messages: 1: In mean.default(x) : l'argument n'est ni numérique, ni logique : renvoi de NA 2: In var(x) : NAs introduced by coercion 3: In mean.default(y) : l'argument n'est ni numérique, ni logique : renvoi de NA 4: In var(y) : Error in if (stderr < 10 * .Machine$double.eps * max(abs(mx), abs(my))) stop("data are essentially constant") :  missing value where TRUE/FALSE needed

Another solution I saw would by to use the gather function to get one column with the Lipids, one column for the value of each Lipids, then create a list column, spread the dataframe and mutate a new-column containing the p-value of the t-test.

tips %>% 
  select(tip, total_bill, sex) %>% 
  gather(key = variable, value = value, -sex) %>% 
  group_by(sex, variable) %>% 
  summarise(value = list(value)) %>% 
  spread(sex, value) %>% 
  group_by(variable) %>% 
  mutate(p_value = t.test(unlist(Female), unlist(Male))$p.value,
         t_value = t.test(unlist(Female), unlist(Male))$statistic)

(https://sebastiansauer.github.io/multiple-t-tests-with-dplyr/)

I'm honestly not sure what to do. Does anyone have tips or anything?

Here's the dput() for the data.... Not really sure why it's necessary though...

dput(dummy)
structure(list(ID = c("A", "A", "A", "A", "A", "A", "A", "A", 
"A", "A", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B"), 
    Lipid.1 = c(0.737, 0.419, 0.468, 0.805, 1.036, 0.825, 0.286, 
    1.166, 0.898, 0.504, 1.433, 0.41, 0.325, 0.866, 0.337, 0.876, 
    0.636, 0.953, 0.481, 0.602), Lipid.2 = c(0.001, 0.017, 0.013, 
    0.025, 0.018, 0.003, 0.007, NA, 0.01, 0.002, 0.01, 0.022, 
    0.005, NA, 0.018, NA, 0.015, 0.016, NA, 0.01), Lipid.3 = c(0.035, 
    0.018, 0.036, 0.024, 0.023, 0.027, 0.036, 0.037, 0.013, 0.037, 
    0.03, 0.04, 0.038, 0.033, 0.016, 0.034, 0.029, 0.033, 0.018, 
    0.029), Lipid.4 = c(NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_), Lipid.5 = c(0.09, 
    0.099, 0.12, 0.058, 0.136, 0.103, 0.153, 0.148, 0.047, 0.085, 
    0.098, 0.133, 0.099, 0.121, 0.084, 0.065, 0.11, 0.088, 0.065, 
    0.043), Lipid.6 = c(0.39, 0.555, 0.568, 0.6, 0.626, 0.378, 
    0.657, 0.57, 0.271, 0.41, 0.474, 0.617, 0.491, 0.738, 0.459, 
    0.365, 0.499, 0.388, 0.271, 0.275), Lipid.7 = c(NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
    NA_real_), Lipid.8 = c(0.186, 0.197, 0.191, 0.125, 0.209, 
    0.107, 0.174, 0.143, 0.055, 0.134, 0.148, 0.193, 0.184, 0.213, 
    0.134, 0.085, 0.165, 0.215, 0.163, 0.061), Lipid.9 = c(NA, 
    NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
    NA, NA, NA, "0,007"), Lipid.10 = c("0,242", "0,254", "0,134", 
    "0,226", "0,243", "0,122", "0,082", "0,119", "0,098", "0,093", 
    "0,27", "0,284", "0,258", "0,236", "0,173", "0,106", "0,138", 
    "0,066", "0,072", "0,081"), Lipid.11 = c("0,053", "0,114", 
    "0,038", "0,094", "0,073", "0,067", "0,028", "0,022", "0,021", 
    "0,05", "0,085", "0,102", "0,122", "0,096", "0,027", "0,03", 
    NA, "0,078", "0,066", NA), Lipid.12 = c(0.223, 0.261, 0.258, 
    0.212, 0.168, 0.101, 0.191, 0.09, 0.195, 0.082, 0.155, 0.2, 
    0.167, 0.231, 0.145, 0.089, 0.239, 0.141, 0.106, 0.124), 
    Lipid.13 = c(0.737, 0.763, 0.707, 0.587, 0.545, 0.317, 0.74, 
    0.602, 0.481, 0.531, 0.632, 0.448, 0.62, 0.766, 0.397, 0.623, 
    0.997, 0.578, 0.418, 0.412), Lipid.14 = c(0.683, 0.666, 0.507, 
    0.366, 0.443, 0.266, 0.493, 0.345, 0.368, 0.355, 0.432, 0.411, 
    0.491, 0.565, 0.357, 0.285, 0.604, 0.426, 0.538, 0.295), 
    Lipid.15 = c(0.911, 1.017, 0.503, 0.76, 0.741, 0.486, 0.648, 
    0.581, 0.955, 0.515, 0.932, 0.707, 0.626, 0.928, 0.836, 0.537, 
    0.654, 0.351, 0.498, 0.529), Lipid.16 = c(0.148, 0.116, 0.069, 
    0.104, 0.091, 0.064, 0.093, 0.123, 0.11, 0.097, 0.283, 0.076, 
    0.095, 0.194, 0.06, 0.061, 0.086, 0.051, 0.064, 0.059), Lipid.17 = c("0,155", 
    "0,274", "0,149", "0,127", "0,174", "nd", "0,109", "0,134", 
    "0,1", "0,09", "0,25", "0,112", "0,088", "0,243", "0,092", 
    "0,073", "0,153", "0,12", "0,14", "0,06"), Lipid.18 = c(3.143, 
    3.441, 4.359, 1.945, 2.573, 2.267, 3.585, 3.405, 2.296, 1.998, 
    3.468, 2.98, 3.626, 3.635, 3.236, 2.092, 2.586, 2.08, 1.718, 
    1.736), Lipid.19 = c(37.993, 36.148, 40.244, 30.395, 37.339, 
    35.742, 47.316, 47.555, 34.351, 32.377, 38.694, 39.413, 36.114, 
    41.235, 32.779, 32.222, 36.418, 36.918, 33.334, 31.421), 
    Lipid.20 = c(6.613, 5.913, 9.662, 3.789, 7.485, 6.297, 8.254, 
    8.07, 4.905, 5.686, 7.742, 7.533, 6.875, 7.908, 7.022, 5.446, 
    6.1, 6.782, 6.062, 6.089), Lipid.21 = c(7.235, 6.759, 8.331, 
    4.931, 6.558, 4.186, 5.99, 5.629, 3.066, 3.439, 7.102, 7.655, 
    6.606, 7.858, 5.804, 3.135, 3.218, 3.639, 2.975, 3.13), Lipid.22 = c(6.453, 
    6.664, 9.048, 4.341, 8.03, 7.599, 10.24, 10.954, 5.873, 6.687, 
    8.005, 8.908, 6.708, 8.06, 5.931, 6.083, 5.734, 5.587, 5.388, 
    6.088), Lipid.23 = c(4.943, 3.164, 5.153, 2.51, 4.071, 5.255, 
    7.636, 8.376, 4.726, 5.56, 4.762, 5.044, 4.549, 4.875, 4.57, 
    5.147, 4.396, 4.031, 3.556, 4.38), Lipid.24 = c(3.973, 4.279, 
    5.928, 3.066, 4.95, 4.667, 7.949, 7.268, 4.948, 3.72, 5.137, 
    5.539, 4.006, 5.276, 3.909, 4.163, 4.954, 5.02, 3.961, 4.201
    ), Lipid.25 = c(7.638, 5.224, 8.417, 3.902, 7.267, 6.007, 
    8.256, 7.457, 4.801, 4.86, 7.581, 8.173, 7.57, 8.591, 7.482, 
    5.091, 5.651, 6.577, 5.415, 5.76), Lipid.26 = c(10.225, 8.293, 
    13.188, 5.607, 10.993, 4.491, 5.767, 5.011, 3.589, 3.145, 
    11.471, 12.183, 9.686, 12.562, 9.697, 3.34, 4.186, 4.485, 
    3.23, 4.229), Lipid.27 = c(5.848, 4.856, 6.503, 3.534, 5.358, 
    8.933, 14.034, 12.806, 7.781, 8.094, 6.765, 6.867, 5.539, 
    7.772, 5.883, 7.832, 8.607, 7.586, 6.628, 7.563), Lipid.28 = c(32.941, 
    30.579, 31.358, 15.861, 30.353, 25.222, 35.662, 34.035, 20.338, 
    24.682, 30.698, 34.024, 31.608, 37.539, 24.901, 20.131, 23.126, 
    30.803, 25.639, 18.935)), class = "data.frame", row.names = c(NA, 
-20L))

Solution

  • If you would like to have the full t-test output, you could just loop over the columns:

    If we start with your df:

    data <- structure(list(ID = c("A", "A", "A", "A", "A", "A", "A", "A", 
    "A", "A", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B"), 
        Lipid.1 = c(0.737, 0.419, 0.468, 0.805, 1.036, 0.825, 0.286, 
        1.166, 0.898, 0.504, 1.433, 0.41, 0.325, 0.866, 0.337, 0.876, 
        0.636, 0.953, 0.481, 0.602), Lipid.2 = c(0.001, 0.017, 0.013, 
        0.025, 0.018, 0.003, 0.007, NA, 0.01, 0.002, 0.01, 0.022, 
        0.005, NA, 0.018, NA, 0.015, 0.016, NA, 0.01), Lipid.3 = c(0.035, 
        0.018, 0.036, 0.024, 0.023, 0.027, 0.036, 0.037, 0.013, 0.037, 
        0.03, 0.04, 0.038, 0.033, 0.016, 0.034, 0.029, 0.033, 0.018, 
        0.029), Lipid.4 = c(NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_), Lipid.5 = c(0.09, 
        0.099, 0.12, 0.058, 0.136, 0.103, 0.153, 0.148, 0.047, 0.085, 
        0.098, 0.133, 0.099, 0.121, 0.084, 0.065, 0.11, 0.088, 0.065, 
        0.043), Lipid.6 = c(0.39, 0.555, 0.568, 0.6, 0.626, 0.378, 
        0.657, 0.57, 0.271, 0.41, 0.474, 0.617, 0.491, 0.738, 0.459, 
        0.365, 0.499, 0.388, 0.271, 0.275), Lipid.7 = c(NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, NA_real_, 
        NA_real_), Lipid.8 = c(0.186, 0.197, 0.191, 0.125, 0.209, 
        0.107, 0.174, 0.143, 0.055, 0.134, 0.148, 0.193, 0.184, 0.213, 
        0.134, 0.085, 0.165, 0.215, 0.163, 0.061), Lipid.9 = c(NA, 
        NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
        NA, NA, NA, "0,007"), Lipid.10 = c("0,242", "0,254", "0,134", 
        "0,226", "0,243", "0,122", "0,082", "0,119", "0,098", "0,093", 
        "0,27", "0,284", "0,258", "0,236", "0,173", "0,106", "0,138", 
        "0,066", "0,072", "0,081"), Lipid.11 = c("0,053", "0,114", 
        "0,038", "0,094", "0,073", "0,067", "0,028", "0,022", "0,021", 
        "0,05", "0,085", "0,102", "0,122", "0,096", "0,027", "0,03", 
        NA, "0,078", "0,066", NA), Lipid.12 = c(0.223, 0.261, 0.258, 
        0.212, 0.168, 0.101, 0.191, 0.09, 0.195, 0.082, 0.155, 0.2, 
        0.167, 0.231, 0.145, 0.089, 0.239, 0.141, 0.106, 0.124), 
        Lipid.13 = c(0.737, 0.763, 0.707, 0.587, 0.545, 0.317, 0.74, 
        0.602, 0.481, 0.531, 0.632, 0.448, 0.62, 0.766, 0.397, 0.623, 
        0.997, 0.578, 0.418, 0.412), Lipid.14 = c(0.683, 0.666, 0.507, 
        0.366, 0.443, 0.266, 0.493, 0.345, 0.368, 0.355, 0.432, 0.411, 
        0.491, 0.565, 0.357, 0.285, 0.604, 0.426, 0.538, 0.295), 
        Lipid.15 = c(0.911, 1.017, 0.503, 0.76, 0.741, 0.486, 0.648, 
        0.581, 0.955, 0.515, 0.932, 0.707, 0.626, 0.928, 0.836, 0.537, 
        0.654, 0.351, 0.498, 0.529), Lipid.16 = c(0.148, 0.116, 0.069, 
        0.104, 0.091, 0.064, 0.093, 0.123, 0.11, 0.097, 0.283, 0.076, 
        0.095, 0.194, 0.06, 0.061, 0.086, 0.051, 0.064, 0.059), Lipid.17 = c("0,155", 
        "0,274", "0,149", "0,127", "0,174", "nd", "0,109", "0,134", 
        "0,1", "0,09", "0,25", "0,112", "0,088", "0,243", "0,092", 
        "0,073", "0,153", "0,12", "0,14", "0,06"), Lipid.18 = c(3.143, 
        3.441, 4.359, 1.945, 2.573, 2.267, 3.585, 3.405, 2.296, 1.998, 
        3.468, 2.98, 3.626, 3.635, 3.236, 2.092, 2.586, 2.08, 1.718, 
        1.736), Lipid.19 = c(37.993, 36.148, 40.244, 30.395, 37.339, 
        35.742, 47.316, 47.555, 34.351, 32.377, 38.694, 39.413, 36.114, 
        41.235, 32.779, 32.222, 36.418, 36.918, 33.334, 31.421), 
        Lipid.20 = c(6.613, 5.913, 9.662, 3.789, 7.485, 6.297, 8.254, 
        8.07, 4.905, 5.686, 7.742, 7.533, 6.875, 7.908, 7.022, 5.446, 
        6.1, 6.782, 6.062, 6.089), Lipid.21 = c(7.235, 6.759, 8.331, 
        4.931, 6.558, 4.186, 5.99, 5.629, 3.066, 3.439, 7.102, 7.655, 
        6.606, 7.858, 5.804, 3.135, 3.218, 3.639, 2.975, 3.13), Lipid.22 = c(6.453, 
        6.664, 9.048, 4.341, 8.03, 7.599, 10.24, 10.954, 5.873, 6.687, 
        8.005, 8.908, 6.708, 8.06, 5.931, 6.083, 5.734, 5.587, 5.388, 
        6.088), Lipid.23 = c(4.943, 3.164, 5.153, 2.51, 4.071, 5.255, 
        7.636, 8.376, 4.726, 5.56, 4.762, 5.044, 4.549, 4.875, 4.57, 
        5.147, 4.396, 4.031, 3.556, 4.38), Lipid.24 = c(3.973, 4.279, 
        5.928, 3.066, 4.95, 4.667, 7.949, 7.268, 4.948, 3.72, 5.137, 
        5.539, 4.006, 5.276, 3.909, 4.163, 4.954, 5.02, 3.961, 4.201
        ), Lipid.25 = c(7.638, 5.224, 8.417, 3.902, 7.267, 6.007, 
        8.256, 7.457, 4.801, 4.86, 7.581, 8.173, 7.57, 8.591, 7.482, 
        5.091, 5.651, 6.577, 5.415, 5.76), Lipid.26 = c(10.225, 8.293, 
        13.188, 5.607, 10.993, 4.491, 5.767, 5.011, 3.589, 3.145, 
        11.471, 12.183, 9.686, 12.562, 9.697, 3.34, 4.186, 4.485, 
        3.23, 4.229), Lipid.27 = c(5.848, 4.856, 6.503, 3.534, 5.358, 
        8.933, 14.034, 12.806, 7.781, 8.094, 6.765, 6.867, 5.539, 
        7.772, 5.883, 7.832, 8.607, 7.586, 6.628, 7.563), Lipid.28 = c(32.941, 
        30.579, 31.358, 15.861, 30.353, 25.222, 35.662, 34.035, 20.338, 
        24.682, 30.698, 34.024, 31.608, 37.539, 24.901, 20.131, 23.126, 
        30.803, 25.639, 18.935)), class = "data.frame", row.names = c(NA, 
    -20L))
    

    clean up a the df:

    # remove the columns which only contain NA:
    data$Lipid.4 <- NULL
    data$Lipid.7 <- NULL
    data$Lipid.9 <- NULL
    
    # convert from string to numeric (I do it now manually with each column. You could use a for-loop)
    data$Lipid.10 <- gsub(",", ".", data$Lipid.10)  # convert comma to dot
    data$Lipid.10 <- as.numeric(data$Lipid.10) # convert from string to numeric
    data$Lipid.11 <- gsub(",", ".", data$Lipid.11)
    data$Lipid.11 <- as.numeric(data$Lipid.11)
    data$Lipid.17 <- gsub(",", ".", data$Lipid.17)
    data$Lipid.17 <- as.numeric(data$Lipid.17)
    
    # get the lipid column names
    all_lipids <- colnames(data)
    all_lipids <-  all_lipids[all_lipids != "ID"] # we don't need the ID column for the loop
    
    # now loop over each column an perform a t-test
    for (column in all_lipids) {
      print(column)
      print(t.test(data[,column] ~ data$ID))
    }
    

    You get for each lipid:

    [1] "Lipid.1"
    
        Welch Two Sample t-test
    
    data:  data[, column] by data$ID
    t = 0.15843, df = 17.391, p-value = 0.8759
    alternative hypothesis: true difference in means is not equal to 0
    95 percent confidence interval:
     -0.2766112  0.3216112
    sample estimates:
    mean in group A mean in group B 
             0.7144          0.6919 
    

    And just a final coment: you perform a lot of comparisons. You may consider to correct for multiple testing.