Dear R users I am currently investigating time series analysis using an irregular time series. Our study is looking at vegetation change in areas of alien vegetation growth after clearing events. The irregular time series is sourced from Landsat ETM+ data, over a six year period I have 38 scenes. For certain periods I have monthly data while for others, images are up to three months apart. So far I have been using linear regression between NDVI (Normalized Difference Vegetation Index) and time to get an overall trend and then plotting the long term trend with the mFilter (Hodrick-Prescott filter). Additional to this I would like to plot a full time series of monthly rainfall data (with no missing data) against my irregular ts. I thus have the following questions, 1. While the mFilter does provide a good trend profile I would like to use a ts analysis procedure which is tailored to irregular environmental data, could anyone suggest a filter / analysis technique besides the mFilter? I am interested in the long term trend, but would also like to identify stochastic changes in the ts? 2. Is it possible to pad / interpolate missing values in a ts and how scientifically robust is this? 3. If interpolation / pad is not an issue how do I deal with NA values in a ts, the mFilter does not like NA and will return only NA if the ts contains any NA's? I am using R version 2.8 on a Dell Precision 690 Workstation running Ubuntu Hardy Heron. If any one has experience with time series analysis and has any suggestions regarding the questions posted, I would really appreciate some help. Many thanks and kind regards, Wesley Wesley Roberts MSc. Researcher: Earth Observation (Ecosystems) Natural Resources and the Environment CSIR Tel: +27 (21) 888-2490 Fax: +27 (21) 888-2693 "To know the road ahead, ask those coming back." - Chinese proverb -- This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard. The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html. This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. MailScanner thanks Transtec Computers for their support.
Ben Fissel
2009-Apr-02 16:33 UTC
[R] [R-sig-Geo] Time series analysis with irregular time-series
Hello Wesely, The appropriate way to address irregular time series depends on what want to use the estimates for. If your objective is to estimate the times that you don't observe (interpolate) then a natural cubic spline is a good method to provide such an estimate. If your objective is to obtain the best one-step (or multi-step) ahead forecast the the Kalman filter is the best method for dealing with irrrgeular time series. Here is a book from the econometrics literature that deals with the subject irregular time series. http://search.barnesandnoble.com/Messy-Data-Missing-Observations-Outliers-and-Mixed-Frequency-Data/R-Carter-Hill/e/9780762303038 With a little work you can probably find something similar in the statistics literature. -Ben On Thu, Apr 2, 2009 at 12:26 AM, Wesley Roberts <wroberts@csir.co.za> wrote:> Dear R users > > I am currently investigating time series analysis using an irregular time > series. Our study is looking at vegetation change in areas of alien > vegetation growth after clearing events. The irregular time series is > sourced from Landsat ETM+ data, over a six year period I have 38 scenes. For > certain periods I have monthly data while for others, images are up to three > months apart. So far I have been using linear regression between NDVI > (Normalized Difference Vegetation Index) and time to get an overall trend > and then plotting the long term trend with the mFilter (Hodrick-Prescott > filter). Additional to this I would like to plot a full time series of > monthly rainfall data (with no missing data) against my irregular ts. I thus > have the following questions, > > 1. While the mFilter does provide a good trend profile I would like to use > a ts analysis procedure which is tailored to irregular environmental data, > could anyone suggest a filter / analysis technique besides the mFilter? I am > interested in the long term trend, but would also like to identify > stochastic changes in the ts? > > 2. Is it possible to pad / interpolate missing values in a ts and how > scientifically robust is this? > > 3. If interpolation / pad is not an issue how do I deal with NA values in a > ts, the mFilter does not like NA and will return only NA if the ts contains > any NA's? > > I am using R version 2.8 on a Dell Precision 690 Workstation running Ubuntu > Hardy Heron. > > If any one has experience with time series analysis and has any suggestions > regarding the questions posted, I would really appreciate some help. > > Many thanks and kind regards, > Wesley > > > > Wesley Roberts MSc. > Researcher: Earth Observation (Ecosystems) > Natural Resources and the Environment > CSIR > Tel: +27 (21) 888-2490 > Fax: +27 (21) 888-2693 > > "To know the road ahead, ask those coming back." > - Chinese proverb > > > > -- > This message is subject to the CSIR's copyright terms and conditions, > e-mail legal notice, and implemented Open Document Format (ODF) standard. > The full disclaimer details can be found at > http://www.csir.co.za/disclaimer.html. > > This message has been scanned for viruses and dangerous content by > MailScanner, > and is believed to be clean. MailScanner thanks Transtec Computers for > their support. > > _______________________________________________ > R-sig-Geo mailing list > R-sig-Geo@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/r-sig-geo >-- Benjamin Earl Fissel Economics Graduate Student University of California, San Diego bfissel.googlepages.com/home bfissel@ucsd.edu bfissel@gmail.com [[alternative HTML version deleted]]