I am attempting to use SAS 9.4 to perform time series analysis so I can understand basic concepts. Despite heroic efforts I am utterly lost. I have literally spent 3 weeks going over hundreds of webpages, books, journals, and articles for well over 200 hours trying to understand the basics and feel no closer to my goal then when I began. I have no idea what the SAS output is indicating and there is no straight answer or tutorial explaining how to read table or chart outputs to let me get on with it. Most of what is out there seems to assume you already know what time series elements are ...things such as level and damping factors ---and I do not. Or they are written by those more interested in showing off programming proweress instead of just creating very simple, easy to follow examples of the different situations that can arise with results. I don't want to create massive datasets and process them. I don't want to extrapolate between dates. That's irrelevant to my situation. Showing me an ACF and PACF plot is great when it deals with nicely lined up data but what if your plots come out like mine do. What then? I just want to know - is my data seasonal? (I don't think so). Is it cyclical? (not on a regular cycle). Is there a trend? (not really). Are there diagnostics I can use on the data to build an appropriate ARIMA model? This work is driving me nuts. Questions: 1. What models should I be using (UCM, ESM, ARIMA?) and what are the inputs I should be using to them? My data does not appear to have any regular cycle that I can see - it varies from 4 to 10 years. Now what? My data does not appear to have seasonality but I wanted to perform Simple Exponential Smoothing with seasonality and vary alpha the alpha in the statement appears to deal with confidence in the solution, not the alpha in this:YHATt|t+1 = alpha*Yt + alpha*(1-alpha)*Yt-1 where can I vary the alpha - it seems to tell me my optimal value (usually either 0.001 or 0.999) but can I manually vary it somewhere? 2. What do the outputs tell me, eg. What does a Spectral Density plot reveal about my data (the cycle length?) Where do I input that into a model? When I get a damping factor calculated, what exactly is it telling me? It's always about 0.88. And??? What's the difference between a standardized ACF / PACF chart and the regular ACF/PACF ones? How do I interpret the ACF and PACF plots? What's IACF? What's a Smoothed Season State plot? What is "Level Weight" and "Seasonal Weight?" ...I assume SW is alpha ideal ~ either approx. 0 or approx. 1? What's the level weight mean? More basic, what is a level? How do you interpret the Final Estimates of the Free Parameters table? What are the meaning of "level" "slope" and "cycle" "damping factor" "period" and "error variance" in Component and Parameters Approx Pr > |t| is "good" when it is small, but what does it mean when it's almost 1? How do you use these results in ARIMA? In the Conditional Least Squares Estimation table, is the Pr>t that is small "good?" ... and therefore AR1,3 and MA1,3 are indicated - and how does that translate into the ARIMA(a,b,c) syntax? Is it a first difference so it's ARIMA(3,1,3)???????? I realize this is a bunch of questions. But I chose SAS and so I'm on my own. My class uses R but I have not because I prefer SAS. I have already spent a lot of money on books that don't help and am just blown away that it's so hard to get straight information on what I need to do to move on with my work. I don't want more phi's or psi's or theories. Just interpretations of actual results! Can someone look at this data and what I've coded so far (attached in zip file) and try to help me along?
... View more
