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.
bin [thumbnails].jpgBin Ren says: My model finally includes local level component and seasonal component. The period of seasonal compenent is 52 weeks, because every year there are periodic transportation peaks which lead to a high demand of gasoline. I use two harmonics in the model, due to the fact in my previous postprevious post that there are overall two peaks for summer and winter. Some rare events such as economic ressesion, are considered as random noise. Based on my model, forecast (95% interval) of regular all formulations retail gasoline price in week 515 (Mar 29 2010 which is 5 week after last observation) and week 540 (Sep 20 2010 which is 30 weeks after last observation) is 319.1046 (285.5028, 356.6612), and 319.7633 (244.3232, 418.4972). Bin_blog2_pic.jpg

blog comments powered by Disqus


17 March 2010