This [unedited] guest post is by a student in my PSTAT262MC class (background post). Please praise/critique/comment on its quality and importance to you.
hamid [thumbnails].jpgHamid Ghofrani says: The temperature of three lakes( lake Michigan, Lake Tahoe and Lake caspian) were presented at the previous post. The multivariate Dynamic Linear Model method was used To model the temperatures for these lakes. Due to the multivariate nature of the data,the SUTSE (Seemingly unrelated Time Series Equations) model including linear trend and seasonal components were used. Within Bayesian framework, Gibbs sampling with wishart priors was used to estimate the unknown parameters (the observation and evolution variance/covariance matrices for the linear trend components). No Covariance structure (both for the observation and evolution) was considered for the seasonal component. The forcasted temperature and their %95 confidence bounds for the one step ahead future month( Jan 2009), all in Celsious are as following . Lake Michigan 8.9(6.15,11.62), Lake Tahoe 11.35(8.74,13.88) and Lake Caspian 15.43(13.4,17.4). Because of very strong seasonality, the deseasend temperature(mean level temperature) of the three lakes for the two months ahead forecasts( January and February of 2009) are shown below. hamid_blog2.jpg

blog comments powered by Disqus


19 March 2010