如何使用分数数据计算可变列数的平均变化

Question

I am trying to find the average difference in a score over repeated measures. The problem is, not every observation is measured unequally often and the values in the columns represent scores on 6 point scale.

the data is present it both Long and wide format like this:

ID    Type    M1    M2    M3    M4    M6
1      A       5     5    3
2      A       4     3    1
3      A       2     5    3     5      5
4      C       5     4    4     3
5      B       3 
6      F       4     2    3     4      1

This is the alternative format:

ID    Type    M    Score
1       A     1      5
1       A     2      5
1       A     3      3
2       A     1      4
2       A     2      3
2       A     3      1
4       C     1      5
4       C     2      4
4       C     3      4
4       C     4      3

I am not really interested in the interim values, but I need the difference between M1 and whatever is the last measurement for that ID then I need to take the average of those differences. I will need to do it across all types and then later broken down by type.

Packages installed are: dplyr, purrr, stringr, tydir, tibble, data.table

The closest I got was the following:

df %>% group_by(M)%>%
    arrange(M)%>%
    summarize(avg = as.numeric(mean(diff(Score))), sd = 
as.numeric(sd(diff(Score))))

and

df %>% group_by(Type)%>%
    arrange(M)%>%
    summarize(avg = as.numeric(mean(diff(Score))), sd = 
as.numeric(sd(diff(Score))))

This was done on the Long format data and gave the result:

       M           avg       sd
     <fctr>       <dbl>    <dbl>
 1            1          NA       NA
 2            2          NA       NA
 3            3 -0.03370787 1.741534
 4            4 -0.04878049 2.036556
 5            5 -0.18181818 1.887760
 6            6  0.00000000 1.095445
 7            7         NaN       NA
 8            8         NaN       NA
 9            9         NaN       NA
10         <NA> -0.16666667 1.722401

The table above is taken from my analysis and not related to example tables. The NA and NaN are a problem as I know there is data in some of the sections, but it is unable to calculate the average difference.

Answer 1

One solution for avg per ID could be using dplyr based on OP feedback to calculate average of difference of first and last measurement.

library(dplyr)

df %>% group_by(ID) %>%
  arrange(M) %>%
  summarise(avg = abs(first(Score) - last(Score))/n())

#Result
#     ID   avg
#  <int> <dbl>
#1     1 0.667
#2     2 1.00 
#3     4 0.500

Actual average and SD for each ID can be calculated as:

df %>% group_by(ID) %>%
  arrange(M) %>%
  summarise(avg = mean(Score), SD = sd(Score))

#Result
     ID   avg    SD
  <int> <dbl> <dbl>
1     1  4.33 1.15 
2     2  2.67 1.53 
3     4  4.00 0.816