arma model fitting

[Bastian Offermann]

Hi,

 

i am trying to fit a simple arma(p,q) model using the fArma package and the
armaFit function therein.

 

 

armaFit(aktie ~ arima(5,0,5), aktie, method = "mle", include.mean = FALSE,
fixed = NULL, description = "ARMA(p,q)")

 

aktie is a numeric vector of length 250 (daily returns of a stock). The
function generates some output, however it also includes

some warning message (see below). Everything ok or am I missing out on
something? Thanks in advance!

 

"Title:

 ARIMA Modelling

 

Call:

 armaFit(formula = aktie ~ arima(5, 0, 5), data = aktie, method = "mle",

    include.mean = FALSE, fixed = NULL, description = "ARMA(p,q) Modell
Optimierung")

 

Model:

 ARIMA(5,0,5) with method: CSS-ML

 

Coefficient(s):

     ar1       ar2       ar3       ar4       ar5       ma1       ma2
ma3       ma4       ma5  

 0.23733   0.08860   0.28411  -0.11786  -0.77622  -0.26711  -0.04569
-0.20281   0.06208   0.79793  

 

Description:

 ARMA(p,q) Modell Optimierung

 

Warning messages:

1: In arima(x = x, order = order, method = method[1], include.mean =
include.mean,  :

  possible convergence problem: optim gave code=1

2: In sqrt(diag(fit$var.coef)) : NaNs have been generated"


        [[alternative HTML version deleted]]

_______________________________________________
R-SIG-Finance@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.

[michal miklovic]

Hi,

the warning message means that the optimisation routine for mle reached the maximum number of iterations before converging. Try increasing the number of iterations by adding the following at the end of the armaFit call:
optim.control = list(maxit = ?)
where ? is the desired number of iterations. The default for ? is 500.
You can also use the 'summary' function to get a more detailed output. Your current model would produce NaNs for some standard errors, see warning no. 2.
As a last remark, I would suggest decreasing the number of ar and ma terms in your model as one of the stylised facts of asset returns is that they exhibit rather low autocorrelation.

Best regards,

Michal




----- Original Message ----
From: Bastian Offermann <bastian2507hk@...>
To: r-sig-finance@...
Sent: Thursday, April 24, 2008 3:47:05 PM
Subject: [R-SIG-Finance] arma model fitting

Hi,



i am trying to fit a simple arma(p,q) model using the fArma package and the
armaFit function therein.





armaFit(aktie ~ arima(5,0,5), aktie, method = "mle", include.mean = FALSE,
fixed = NULL, description = "ARMA(p,q)")



aktie is a numeric vector of length 250 (daily returns of a stock). The
function generates some output, however it also includes

some warning message (see below). Everything ok or am I missing out on
something? Thanks in advance!



"Title:

ARIMA Modelling



Call:

armaFit(formula = aktie ~ arima(5, 0, 5), data = aktie, method = "mle",

    include.mean = FALSE, fixed = NULL, description = "ARMA(p,q) Modell
Optimierung")



Model:

ARIMA(5,0,5) with method: CSS-ML



Coefficient(s):

     ar1       ar2       ar3       ar4       ar5       ma1       ma2
ma3       ma4       ma5  

0.23733   0.08860   0.28411  -0.11786  -0.77622  -0.26711  -0.04569
-0.20281   0.06208   0.79793  



Description:

ARMA(p,q) Modell Optimierung



Warning messages:

1: In arima(x = x, order = order, method = method[1], include.mean =
include.mean,  :

  possible convergence problem: optim gave code=1

2: In sqrt(diag(fit$var.coef)) : NaNs have been generated"


    [[alternative HTML version deleted]]

_______________________________________________
R-SIG-Finance@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.



      ____________________________________________________________________________________

[[elided Yahoo spam]]
        [[alternative HTML version deleted]]

_______________________________________________
R-SIG-Finance@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.