How to make a graph that shows reliable and clinically significant change in particular variable?

[Huy Nguyen]

Dear Statalisters,

I appreciate if anyone can help with how to make a graph showing reliable and clinically significant change in particular variable, something like panel A or B of figure 2 below from a paper. I searched every Stata forum questions and Stata materials related to this plot, including a book on A visual Guide to Stata Graphics (Third Edition) by Michael N. Mitchell, yet nothing matched my question.

I also have quite similar data to the above study, with the following variables as pasted below:



- ID is participant ID, with 99 participants (1-99).
- GAD_0 is depression score at baseline (0-21)
- GAD_8 is depression score at week 8 (post-intervention time) after clinical intervention (0-20)
- Dif_GAD_0_8 is difference score between pre- and post-intervention (-7-20)
- RCCriterion_0_8=4.45312 is the cutoff for assessing the reliable change or reliable improvement. This parameter was estimated based on the formula: 1.96*SE_diff = 1.96*SD* sqrt(2-r1-r2)where SE_diff = SE of the difference; SD = SD of test 1; r1 and r2 are reliability of tests 1 and 2. So, if Dif_GAD_0_8 > 1.96*SE_diff or >4.45312, then we can identify him/her as reliable improvers.
- Actual_Dif_GAD_0_8 is a dichotomous variable classifying a participant as reliable changer or not based on the above criteria.
- A clinical cutoff for GAD defined from literature is a score of 13.

So, with the above data, how can I make a plot of figure 2 (like Panel A) as above, if possible, can you suggest with Stata codes based on Stata data below.

* Example generated by -dataex-. For more info, type help dataex
clear
input float ID double(GAD_0 GAD_8) float(Dif_GAD_0_8 RCCriterion_GAD_0_8 Actual_Dif_GAD_0_8)
 1 20  4 16 4.45312 1
 2  4  7 -3 4.45312 0
 3  9  4  5 4.45312 1
 4  7  6  1 4.45312 0
 5 17  6 11 4.45312 1
 6 18  7 11 4.45312 1
 7  7  4  3 4.45312 0
 8 13  5  8 4.45312 1
 9 12  4  8 4.45312 1
10 18  6 12 4.45312 1
11  7  1  6 4.45312 1
12 14  0 14 4.45312 1
13 12  4  8 4.45312 1
14 17  5 12 4.45312 1
15 12  7  5 4.45312 1
16 21 10 11 4.45312 1
17  3  2  1 4.45312 0
18  4  5 -1 4.45312 0
19  7  5  2 4.45312 0
20 17  4 13 4.45312 1
21  5  4  1 4.45312 0
22 14  5  9 4.45312 1
23  0  3 -3 4.45312 0
24  8 10 -2 4.45312 0
25 12  3  9 4.45312 1
26 11  9  2 4.45312 0
27 15  3 12 4.45312 1
28 21 14  7 4.45312 1
29 16  4 12 4.45312 1
30  4  2  2 4.45312 0
31  8  4  4 4.45312 0
32  7  5  2 4.45312 0
33 10  6  4 4.45312 0
34 16 10  6 4.45312 1
35 12  3  9 4.45312 1
36 14  6  8 4.45312 1
37  4  4  0 4.45312 0
38  6  5  1 4.45312 0
39 20  8 12 4.45312 1
40 15  4 11 4.45312 1
41 10  5  5 4.45312 1
42 16  4 12 4.45312 1
43  4  1  3 4.45312 0
44 19  5 14 4.45312 1
45  9  2  7 4.45312 1
46  6  4  2 4.45312 0
47 12  7  5 4.45312 1
48 14  5  9 4.45312 1
49  7  7  0 4.45312 0
50 16  7  9 4.45312 1
51  9  3  6 4.45312 1
52  1  3 -2 4.45312 0
53  7  4  3 4.45312 0
54 20  8 12 4.45312 1
55 15  4 11 4.45312 1
56  8  6  2 4.45312 0
57 13  6  7 4.45312 1
58 18  7 11 4.45312 1
59 12  4  8 4.45312 1
60 16  6 10 4.45312 1
61  5  4  1 4.45312 0
62 20  0 20 4.45312 1
63  7  5  2 4.45312 0
64 18  6 12 4.45312 1
65 16 12  4 4.45312 0
66 19  0 19 4.45312 1
67 15  8  7 4.45312 1
68  7  4  3 4.45312 0
69 20  7 13 4.45312 1
70 10  3  7 4.45312 1
71 17 20 -3 4.45312 0
72  7  9 -2 4.45312 0
73 14  6  8 4.45312 1
74 11  7  4 4.45312 0
75  9  2  7 4.45312 1
76  7 11 -4 4.45312 0
77 21  9 12 4.45312 1
78  3  3  0 4.45312 0
79 14  9  5 4.45312 1
80 19  0 19 4.45312 1
81 16 13  3 4.45312 0
82  9 16 -7 4.45312 0
83  8  3  5 4.45312 1
84  6  4  2 4.45312 0
85 12  8  4 4.45312 0
86  6  5  1 4.45312 0
87 18  4 14 4.45312 1
88 21  9 12 4.45312 1
89  7  5  2 4.45312 0
90  5  4  1 4.45312 0
91  7  5  2 4.45312 0
92  4  5 -1 4.45312 0
93 20 19  1 4.45312 0
94  1  8 -7 4.45312 0
95 21  6 15 4.45312 1
96 13  8  5 4.45312 1
97  3  4 -1 4.45312 0
98 11  5  6 4.45312 1
99 12  5  7 4.45312 1
end

I thank you for your advice and help.

[Maarten Buis]

set scheme s1mono
label define group_lb 1 `""no reliable or" "clinically significant change""' ///
                      2 `""reliable but" "not clinically significant change""' ///
                      3 `""reliable and" "clinically significant change""' ///
                      4 `""not reliable but" "clinically significant change""'
gen byte group:group_lb = 1 if GAD_8 <  13 & Dif_GAD_0_8 <  4.45312
replace group           = 2 if GAD_8 <  13 & Dif_GAD_0_8 >= 4.45312
replace group           = 3 if GAD_8 >= 13 & Dif_GAD_0_8 >= 4.45312
replace group           = 4 if GAD_8 >= 13 & Dif_GAD_0_8 <  4.45312

separate GAD_0, by(group) veryshortlabel

twoway scatter GAD_0? GAD_8,             ///
    aspect(1)                            ///
    legend(col(1) pos(4) order(1 2 3 4)) ///
    xline(13) yline(13)               || ///
    function y=x, range(0 21)

[Huy Nguyen]

Thank you so much Maarten for your excellent codes. It is super-useful.