interrupted time series/ segmented regression

esraa1 — Thu, 19 Mar 2020 19:30:58 GMT

hi!
I have  a data ( number of  crashes ). I used interrupted time series to show the effect of speed camera that introduced at time=49
I want to plot in one graph as in the Figure1a in the link http://www.cdc.gov/pcd/issues/2013/12_0268.htm 

30 1 
26 2
39 3 
44 4 
35 5     
20 6 
27 7 
67 8 
25 9 
39 10 
69 11   
74 12 
45 13 
41 14 
50 15 
67 16 
55 17   
77 18 
41 19 
38 20
38 21 
46 22 
45 23   
62 24  
60 25  
62 26 
49 27 
60 28
78 29         
69 30      
64 31 
86 32
68 33 
53 34 
59 35 
56 36 
68 37 
57 38 
59 39 
101 40 
63 41
90 42 
92 43 
64 44      
77 45 
69 46
82 47 
70 48 
63 49 
66 50  
78 51 
94 52 
88 53 
62 54 
77 55 
92 56 
76 57 
55 58 
70 59 
54 60 
76 61 
58 62 
58 63
79 64 
77 65 
79 66 
;

Re: interrupted time series/ segmented regression

RichardDeVen — Sat, 21 Mar 2020 06:37:27 GMT

Create two additional variables that mimic the original data, for use in SGPLOT REG statements.

data have;
input crash_count time;
if time <  49 then y1 = crash_count;  * y1 missing when time >=49;
if time >= 49 then y2 = crash_count; * y2 missing when time < 49;
datalines;
30 1
26 2
39 3
44 4
35 5
etc...


proc sgplot data=have;
  series x=time y=crash_count / markers markerattrs=(symbol=circlefilled) ;
  reg    x=time y=y1 / nomarkers lineattrs=(pattern=dot  color=black) legendlabel='Trendline Pre-49';
  reg    x=time y=y2 / nomarkers lineattrs=(pattern=dash color=black) legendlabel='Trendline Post-49';
  keylegend / across=1 location=inside;
run;

Note: The REG statement has no option for extending a fit line (such as the pre-49 dotted line) from left edge to right edge.

topic interrupted time series/ segmented regression in SAS Programming

interrupted time series/ segmented regression

Re: interrupted time series/ segmented regression