12-10-2014 01:44 PM
Suppose I have a collection of sets of numbers, lets say that each set has twenty five numbers in it (one for each day in a twenty five day period). Lets say I have 200 of these sets.
Are there any statistical methods which will help me to identify trends in the data which can predict (based on my existing sets) whether a new sets number on the 25th day day will be lower than on the first day?
For instance, suppose I am on the fifth day. Can I calculate a percentage chance that my last day will be lower than the first based on the number of observations I have so far?
Any guidance on methods that would point me in the right direction would be much appreciated. I'm using base SAS to analyze my datasets.
Edit: To make the problem more clear ...
I have a dataset with 200 rows and 25 columns. Each row represents time series data for a stock that was added to the S&P 500 index. Each column represents the nth day the price was recorded for. So if row one represents Whole Foods then row one column five represents the price for whole foods on the fifth day after it was added to the index.
Suppose a new company is added to the index. Each day a price is recorded for the new stock. Each day I want to look at all of the days available to me for the new stock and make a prediction about whether the last days price (column 25) will be lower than the first days price. I want to make that prediction based on the whatever trends exist in the 200 rows of historical data I have in my dataset.
Based on some investigation it looks like an autoregression problem for which SAS has proc autoreg. Any experience with that procedure and how it might apply to my particular problem would be appreciated.
Ultimately for each day of existence for the new stock in the index I would like to have a percentage chance that the last day will be lower than the first.
Message was edited by: Brian Duffy