Analyzing two time points

[Katie Holzer]

I am analyzing data with the mental health component score (measured 0-100) for 4500 individuals. This score was measured at baseline and then 30 days following an intervention. Is there any way to identify the number of individuals whose score decreased, increased, or had no change?

[David Benson]

I assumed your data were in a "long format"

*** I created some toy data
dataex id time score // To install: ssc install dataex
clear
input byte(id time score)
1 1 21
1 2 95
2 1 38
2 2  4
3 1 98
3 2 98
4 1 50
4 2 12
5 1 33
5 2 91
6 1 21
6 2 52
end

bysort id (time): gen wanted = "Increased" if score[2] > score[1] & score[2]!=.
bysort id (time): replace wanted = "No change" if score[2] == score[1] & score[2]!=.
bysort id (time): replace wanted = "Decreased" if score[2] < score[1] & score[2]!=.
gen d_increased = (wanted=="Increased")  // 1 if wanted=="Increased", 0 otherwise

. list, sepby(id) abbrev(12) noobs

  +---------------------------------------------+
  | id   time   score      wanted   d_increased |
  |---------------------------------------------|
  |  1      1      21   Increased             1 |
  |  1      2      95   Increased             1 |
  |---------------------------------------------|
  |  2      1      38   Decreased             0 |
  |  2      2       4   Decreased             0 |
  |---------------------------------------------|
  |  3      1      98   No change             0 |
  |  3      2      98   No change             0 |
  |---------------------------------------------|
  |  4      1      50   Decreased             0 |
  |  4      2      12   Decreased             0 |
  |---------------------------------------------|
  |  5      1      33   Increased             1 |
  |  5      2      91   Increased             1 |
  |---------------------------------------------|
  |  6      1      21   Increased             1 |
  |  6      2      52   Increased             1 |
  +---------------------------------------------+

// NOTE: The above puts "Increased", "Decreased", etc in both observations.  If you only want one obs per person marked
bysort id (time): replace wanted = "" if _n==1
bysort id (time): replace d_increased  = . if _n==1

. list, sepby(id) abbrev(12) noobs

  +---------------------------------------------+
  | id   time   score      wanted   d_increased |
  |---------------------------------------------|
  |  1      1      21                         . |
  |  1      2      95   Increased             1 |
  |---------------------------------------------|
  |  2      1      38                         . |
  |  2      2       4   Decreased             0 |
  |---------------------------------------------|
  |  3      1      98                         . |
  |  3      2      98   No change             0 |
  |---------------------------------------------|
  |  4      1      50                         . |
  |  4      2      12   Decreased             0 |
  |---------------------------------------------|
  |  5      1      33                         . |
  |  5      2      91   Increased             1 |
  |---------------------------------------------|
  |  6      1      21                         . |
  |  6      2      52   Increased             1 |
  +---------------------------------------------+


. tabulate wanted

     wanted |      Freq.     Percent        Cum.
------------+-----------------------------------
  Decreased |          2       33.33       33.33
  Increased |          3       50.00       83.33
  No change |          1       16.67      100.00
------------+-----------------------------------
      Total |          6      100.00

[Katie Holzer]

Thank you for answering my question!

[Nick Cox]

Here is some more technique. I stole david Benson's data example and some of his ideas from his helpful post.

nput byte(id time score)
1 1 21
1 2 95
2 1 38
2 2  4
3 1 98
3 2 98
4 1 50
4 2 12
5 1 33
5 2 91
6 1 21
6 2 52
end

bysort id (time) : gen previous = score[1] if _n == 2 

scatter score previous 

gen change = score - previous 
gen sign = sign(change)

label def sign -1 "Down" 0 "Same" 1 "Up"
label val sign sign 

tab sign 

     sign |      Freq.     Percent        Cum.
------------+-----------------------------------
       Down |          2       33.33       33.33
       Same |          1       16.67       50.00
         Up |          3       50.00      100.00
------------+-----------------------------------
      Total |          6      100.00

The same information: but the values -1 0 1 ensure the right order: alphabetical order doesn't do quite the right thing in David's table.

[Katie Holzer]

This is helpful, thank you.