ARIMA Modelling with Binary Independent Variable

[Tom van Oosterhout]

I am trying to fit an ARIMA model to my intervention analysis but when I use the syntax and add both a dependent and independent variable in the command, then it says my binary independent variable has an unknown weight type. Even when I try to add iweight, it still does not work?

Does anyone have any suggestions of how to run an ARIMA model with both a dependent variable and a binary independent variable?

Thanks

[Justin Niakamal]

You can look into the user written Intervention time-series models command tstf (from SSC - type ssc install tstf in Stata). You can run an arima with an intervention date.

For more details see
https://www.stata.com/meeting/nordic...tic17_Zhou.pdf

Hope this helps,
Justin

[Tom van Oosterhout]

Thank you Justin!

I am receiving "weights not allowed" from stata

This is a dataex of my data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input double OverallInflation float n
26.7   1
26.8   2
25.5   3
25.2   4
22.6   5
23.2   6
22.6   7
25.8   8
23.4   9
20.2  10
17.5  11
15.5  12
15.5  13
16.2  14
17.2  15
16.1  16
15.4  17
15.6  18
14.5  19
15.1  20
15.4  21
12.9  22
  12  23
11.5  24
 9.6  25
 9.8  26
10.2  27
 9.8  28
   9  29
 8.5  30
 8.7  31
   9  32
 9.2  33
10.1  34
10.2  35
10.7  36
11.1  37
11.3  38
11.6  39
11.3  40
10.9  41
12.2  42
12.8  43
13.7  44
  14  45
14.2  46
14.9  47
15.3  48
15.5  49
15.9  50
15.6  51
15.4  52
15.8  53
15.9  54
16.2  55
16.5  56
16.7  57
17.1  58
16.6  59
16.1  60
15.8  61
15.3  62
14.9  63
  12  64
11.6  65
  11  66
10.5  67
10.1  68
 9.6  69
 9.2  70
 8.6  71
 8.4  72
 7.9  73
 7.7  74
 7.4  75
 7.2  76
 7.1  77
 7.2  78
 7.4  79
 7.5  80
 7.7  81
   8  82
 8.2  83
 8.1  84
 7.9  85
 8.5  86
 8.7  87
 9.1  88
 9.3  89
 9.4  90
 9.6  91
 9.9  92
10.1  93
 9.7  94
 9.5  95
 9.2  96
 8.7  97
 8.4  98
   8  99
 7.8 100
end

I want to do a tstf intervention analysis with an arima model(3,1,0) and a intervention date (n=85)

I hope you can help me solve this problem

Thanks

[Justin Niakamal]

Try

tsset n
gen intervention = (n >= 85)
arima OverallInflation intervention, arima(3,1,0)

You could also make intervention = 1 in period 85 with

gen intervention = (n == 85)

Normally you'll want to use factor notation, but arima doesn't support it. So just have dummy as regressor.
Hope this helps

[Tom van Oosterhout]

Thank you very much!