[R] Checking whether a time series is stationary with irregular spaced data

[Worthington, Thomas A]

Hi Mark 

Thank you for the response. The problem is biologically I can’t really avoid using the irregular time step (daily samples taken at irregular intervals)

I guess generally I’m not really doing time series analysis (my model  is a multiple regression), I just wanted to use time series techniques to test for stationarity. 

Therefore would it be feasible to aggregate the data to say a weekly time step which is regularly spaced, use one of the methods to check were the data is stationary, then use the irregular daily time step data in the multiple regression model?

Thanks again

Tom      

    

From: Mark Leeds [mailto:markleeds2 at gmail.com] 
Sent: Thursday, July 03, 2014 12:09 AM
To: Worthington, Thomas A
Subject: Re: [R] Checking whether a time series is stationary with irregular spaced data

hi thomas: I don't deal with irregularly spaced data but the standard time series techniques
don't apply there. there may be other ways but one workaround is to interpolate so
that it's regularly spaced and then use that regularly spaced series.

On Wed, Jul 2, 2014 at 6:26 PM, Worthington, Thomas A <thomas.worthington at okstate.edu> wrote:
I attempting to model the relationship between water temperature and air temperature. The seasonal component of the temperature time series has been modeled using a sinusoidal function, leaving the air and water temperature residuals. I want to model the relationship with

M5<- gls(Water ~ Air +Air1 +Air2, correlation = corCAR1(form =~ Date))

Where Water is the water temperature residual, Air is the air temperature residuals at 1 and 2 day lags. I have included an autocorrelation structure that takes into account the fact that the water temperature were taken at irregular spaced intervals.

I would like to test whether the time series is stationary, I found a blog post that used the following graphical methods and tests (Cent_Water is the water temperature centered by subtracting the mean value)

Acf(Cent_Water)
Pacf(Cent_Water)
Box.test(Cent_Water, lag=20, type="Ljung-Box")
adf.test(Cent_Water, alternative="stationary")
kpss.test(Cent_Water)

Are these methods useable with irregular spaced data as I believe it is not possible to use Acf?

Any suggestions would be greatly appreciated

Best wishes
Tom

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.