Hi!
I am trying to use Interrupted time series with ARIMA model to compare before and after at intervention=45
data outcome;
input outcome time intervention time_af_int;
datalines;
9 1 0 0
9 2 0 0
10 3 0 0
8 4 0 0
8 5 0 0
6 6 0 0
6 7 0 0
13 8 0 0
20 9 0 0
23 10 0 0
29 11 0 0
34 12 0 0
19 13 0 0
39 14 0 0
44 15 0 0
29 16 0 0
34 17 0 0
62 18 0 0
50 19 0 0
46 20 0 0
51 21 0 0
36 22 0 0
42 23 0 0
48 24 0 0
30 25 0 0
64 26 0 0
66 27 0 0
77 28 0 0
54 29 0 0
74 30 0 0
48 31 0 0
52 32 0 0
73 33 0 0
77 34 0 0
83 35 0 0
55 36 0 0
48 37 0 0
48 38 0 0
47 39 0 0
44 40 0 0
49 41 0 0
64 42 0 0
35 43 1 1
77 44 1 2
46 45 1 3
58 46 1 4
55 47 1 5
70 48 1 6
41 49 1 7
56 50 1 8
45 51 1 9
57 52 1 10
62 53 1 11
51 54 1 12
76 55 1 13
58 56 1 14
46 57 1 15
71 58 1 16
62 59 1 17
64 60 1 18
59 61 1 19
54 62 1 20
70 63 1 21
54 64 1 22
65 65 1 23
52 66 1 24
56 67 1 25
70 68 1 26
71 69 1 27
70 70 1 28
60 71 1 29
;
run;
there are the steps I went through
1 check for stationarity by using PROC ARIMA
proc arima data=sample;
identify var=outcome stationarity=(adf);
run;
the results showed the outcome is stationarity
2. from ACF PACF plots the AR=2
my questions are
1.how I can estimate the coefficient of the model (b0,b1,b2,b3)with AR=2 if I used the linear regression
outcome=b0+b1*time+b2*intervention+b3*time_af_int
2. can I use the nonlinear regression with ARIMA?
3. how account for the seasonality?