How to use o2fx:
For the tutorial, watch a video.
Collecting Data:
For o2fx, you will most likely want to use time series data, this means data that has been collected for a certain period of time, at discreet time intervals. For example: the population of a town is measured every year at the same time.
There are many places to look for data on the internet. These might be good places to start:
- Yahoo Finance - Stock price histories are available for most stocks. Click here for a quick walkthrough.
- Data.gov - This is the United States Government's public database. Here, you can find all kinds of time series data in many different fields. Of particular interest may be human populations over time.
- Reddit Datasets - This has user uploaded or linked datasets, and discussions about them. This might be a good place to learn more about time series data, which is what we use in o2fx.
Preparing Data:
The key to getting useful information out of o2fx, as for any other tool, is making sure the data you put in is meaningful and relevant to the information you want to receive. Here are some hints:
- Align your Independent (x) and Dependent (y) Inputs - Make sure that they start at the same moment and were sampled at the same interval: If your X data starts on May 5, 2012, and the data was taken every day, then make sure your Y data starts on the same day and was taken every day as well.
- Ensure the correct direction of your data - If you are interested in finding how X predicts Y over time, then you need to make sure that your data starts in the past and moves forward through time. You can use the reverse buttons on the data input to switch the direction of either or both inputs.
- Use space eparated or csv data - If you are copying data from columns or rows in excel, then do not click the csv button because your data is space separated. CSV means Comma Separated Values, and means that each item in the input is separated by a comma, this is common when downloading database information from the web.
- Pick relevant sample sizes - More data doesn't necessarily mean more accuracy. Relationships can change over time, meaning that correlations for 1000 samples of data may be different than for the last 100 samples. But, you may only be interested in how the last 100 samples work together.
- For stocks, price change in dollars may not be as meaningful as price change as a percentage of the stock price. Click here for a walk through on finding the % change.
Additional help: