Re: [pystatsmodels] Statsmodels seasonal ARIMA: only monthly and quarterly periods are supported

[josef...@gmail.com]

On Wed, Aug 19, 2015 at 7:45 AM, Maha Mdini <medin...@gmail.com> wrote:

Hello,

I am implementing seasonal ARIMA in Python and I am using Statsmodels 0.7.0. I am calling the function statsmodels.tsa.x13.x13_arima_select_order with a 15 minutes periodic data in argument. I am getting this error:

"Only monthly and quarterly periods are supported. Please report or send a pull request if you want this extended."

I do not find a way to apply seasonal ARIMA to data with different periods. Would you please help me solve this error?

Thanks in advance for your time.

I guess this is a hardcoded linimtation of x13, because it was written for macroeconomic data. You could check the documentation for it, e.g. https://www.census.gov/srd/www/winx13/

statsmodels master also has SARIMAX since last Winter, which doesn't impose a restriction on the period length. However, very long seasonality will require large statespace matrices, AFAIK.

I don't think there is a sarimax_select_order support yet.

Josef

 

Best regards,

[Nicolas Gutierrez]

Hi guys,

    I've also ran into something similar.
    Did you find out how set a custom period in x13?
    For instance, I can do this in statsmodels.tsa.seasonal seasonal_decompose
    simply by using the "freq" argument.
    i.e. seasonal_decompose( .., freq = custom_period)

     BTW has anyone compared seasonal_decompose and x13?
     Any idea what the strengths/weaknesses are for each algorithm?

     Best,
        Nicolas

On Wednesday, August 19, 2015 at 1:57:52 PM UTC+2, Maha Mdini wrote:

Hello,

I am implementing seasonal ARIMA in Python and I am using Statsmodels 0.7.0. I am calling the function statsmodels.tsa.x13.x13_arima_select_order with a 15 minutes periodic data in argument. I am getting this error:

"Only monthly and quarterly periods are supported. Please report or send a pull request if you want this extended."

I do not find a way to apply seasonal ARIMA to data with different periods. Would you please help me solve this error?

Thanks in advance for your time.

Best regards,