# How to I compare the growth differences between two time based datasets?

(See note**)

I am attempting to put together an infograph and am almost done collecting all the data I need.

In addition to the base raw data collected, I'd like to add something I'm calling parallel stability and differential volatility. Here's what I mean:

• First, take a look at these graphs:

• While interesting, I'd like to add data for growth/recession realized between the two parties for a trailing 14 days.
• Also like to be able to add the "parallel stability and differential volatility" measurement, which basically says if both party experience the same growth/recession during the same time period they have "parallel stability". On the other hand, if one party grows, and the other experiences a recess in growth, that is differential volatility.

I'm looking to compare YouTube-to-Youtube, Searches-to-Searches, News-to-News data between the two parties, not across the datasets.

If it's of use I can attempt to post the data, but it's basically one data point per day for the past 365 days per dataset (YouTube, Searches, News) per person with Y-values between 1-100; meaning you could just as easily create a dummy dataset based on how you describe how to find the solution.

** Let me clearly state that my understanding of math, stats, etc - is very limited; so assume that I know almost nothing and use plain-English where possible.

-
What do you mean by growth/recession. Why not just compute the 14 day percent change in your indicator and the relative difference between the two? Fancy names tend to make things more confusing. –  John Aug 14 '12 at 16:52
+1 @John: Agree about the fancy names. I understand how to calculate the 14-day percent change, what would be the calculation for the relative difference? –  blunders Aug 14 '12 at 17:09
From your description, it sounds like you want the changes for D's less the changes for R's (simple subtraction). If the result is 0, you have stability; if its greater than 0 you have volatility. Am I missing something? –  dav Aug 14 '12 at 17:20
@David Vandenbos: Yes, that makes sense. Should I just delete the question, sense I'm guessing based on the way I asked the question, it's not adding value to the site. Thanks! –  blunders Aug 14 '12 at 17:31
Why not finish working out your answer and then writing it up as an answer to your question-that way the site benefits from the exchange. The community/mods will let you know if the question isn't a good fit and should be deleted-but usually you'd have gotten a sense of that by now. –  dav Aug 14 '12 at 18:13

First, calculate the growth rate percentage over a single span of time for each point of calculation, in this case for everyday in the dataset but the first 14-days. To do this simply:

• Determine the value of the day 14 days before the day being evaluated and for the day being evaluated. For example, let's say the day being evaluated has a value of 60, and the day 14-days before it has a value of 20.
• Next subtract the value of the day 14-days before the day before evaluated from the day evaluated. For the example, your calculation would be 60 - 20 = 40.
• Divide the difference by the day 14-days before the day before evaluated. In this case, 40/60 = .6666.
• Lastly multiply the result from the last step by 100 to change it to a percentage: .6666 x 100 = 66.66. Meaning the day being evaluated had a growth of 66.66 percent in over the 14-days before it.

Next to compare the difference in growth between two measurements of the same type over the same period of time:

• First complete the growth rate percentage for each point over the timespan.
• Next subtract the difference between the two calculations,
• Lastly depending on how you want to display the data:
• Convert the result to a positive number if you want a single measure of stability/volatility, where zero is the most stable value and the highest value is the most volatile.
• Convert the result to a positive number if you want a single measure of stability/volatility, but unlike in the step take the value being subtracted from and assign all the positive result to it, then assign all the negative results to the figured used to subtract with and convert them to positive values.
-