2
$\begingroup$

In my studies I've been working recently on dependency between debt and GDP growth in USA from 1966 to 2015. I used logged and differenciated GDP time series data and combined it with 0/1 debt-to-GDP time series data which takes value 0 when the debt-to-GDP level is below some arbitrary value and 1 when it's above. I have made 41 such arrays, depending on arbitrary level, and tried to estimate ARIMA model through auto.arima method in "forecast" library and then I'd like to select the best model selecting the best information criterion (let's say Akaike).

However, using auto.arima on series returned with No suitable ARIMA model found. Could somebody tell what is the reason for it and how can I estimate it?

Here is the data I use:

gdp time series:

Time Series:
Start = 1966 
End = 2015 
Frequency = 1 

               gdp
 [1,] 8.150000e+11
 [2,] 8.617000e+11
 [3,] 9.425000e+11
 [4,] 1.019900e+12
 [5,] 1.075884e+12
 [6,] 1.167770e+12
 [7,] 1.282449e+12
 [8,] 1.428549e+12
 [9,] 1.548825e+12
[10,] 1.688923e+12
[11,] 1.877587e+12
[12,] 2.085951e+12
[13,] 2.356571e+12
[14,] 2.632143e+12
[15,] 2.862505e+12
[16,] 3.210956e+12
[17,] 3.344991e+12
[18,] 3.638137e+12
[19,] 4.040693e+12
[20,] 4.346734e+12
[21,] 4.590155e+12
[22,] 4.870217e+12
[23,] 5.252629e+12
[24,] 5.657693e+12
[25,] 5.979589e+12
[26,] 6.174043e+12
[27,] 6.539299e+12
[28,] 6.878718e+12
[29,] 7.308755e+12
[30,] 7.664060e+12
[31,] 8.100201e+12
[32,] 8.608515e+12
[33,] 9.089168e+12
[34,] 9.660624e+12
[35,] 1.028478e+13
[36,] 1.062182e+13
[37,] 1.097751e+13
[38,] 1.151067e+13
[39,] 1.227493e+13
[40,] 1.309373e+13
[41,] 1.385589e+13
[42,] 1.447764e+13
[43,] 1.471858e+13
[44,] 1.441874e+13
[45,] 1.496437e+13
[46,] 1.551793e+13
[47,] 1.615526e+13
[48,] 1.669152e+13
[49,] 1.739310e+13
[50,] 1.803665e+13

debt (50% threshold):

     debt%gdp
 [1,]        0
 [2,]        0
 [3,]        0
 [4,]        0
 [5,]        0
 [6,]        0
 [7,]        0
 [8,]        0
 [9,]        0
[10,]        0
[11,]        0
[12,]        0
[13,]        0
[14,]        0
[15,]        0
[16,]        0
[17,]        0
[18,]        0
[19,]        0
[20,]        0
[21,]        0
[22,]        0
[23,]        0
[24,]        1
[25,]        1
[26,]        1
[27,]        1
[28,]        1
[29,]        1
[30,]        1
[31,]        1
[32,]        1
[33,]        1
[34,]        1
[35,]        1
[36,]        1
[37,]        1
[38,]        1
[39,]        1
[40,]        1
[41,]        1
[42,]        1
[43,]        1
[44,]        1
[45,]        1
[46,]        1
[47,]        1
[48,]        1
[49,]        1
[50,]        1

Then I intersect them and put into auto.arima model.

$\endgroup$
  • 3
    $\begingroup$ I don't know what you mean by "intersect", but you should pass your debt-to-GDP ratio variable as the xreg argument. Aside from that, you may be getting errors because of numerical overflow. I would try using GDP in billions of USD, not in USD. $\endgroup$ – Chris Haug Feb 2 '17 at 13:31
  • 2
    $\begingroup$ This isn't really clear (& won't be on topic on Stack Overflow, because it isn't really a programming question). But I think you are trying to find an ARIMA model for a set of 0s followed by a set of 1s. That isn't the kind of data a standard ARIMA is supposed to work with, so it isn't surprising it complains. $\endgroup$ – gung - Reinstate Monica Feb 2 '17 at 16:04
1
$\begingroup$

I encountered a similar problem when I was trying to run auto.arima with one of my explanatory variables. I took the square root of the variable and auto.arima worked just fine. The problem might be that the values in your GDP variable might be too large and might be creating a numerical overflow.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ This is the correct answer. (+1) If the response variable is sqrted, log10ed or even simply divided by 1000, auto.arima works "fine". $\endgroup$ – usεr11852 Jun 17 '18 at 10:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.