Friday, April 29, 2016

Disorganized Writer Offers Follow-Up to 'Scatterplot Trends' Story

On the 27th I took a scatterplot -- the private-debt-to-public-debt (P2P) ratio versus inflation-adjusted GDP growth -- and tweaked the hell out of it. Ended up with the data not chronological but sorted on the P2P values, with both the x and y values expressed as moving averages of the sorted data, and with a series of short trendlines describing the path of the scatterplot trend.

That was fun to do. But the conclusion I came to, after all the fun, was

Not sure about sorting the values.

Not much of a conclusion, is it? I want to go back and get sure about sorting the values. How would it look if I put the data back in chronological order? That was my question.

Here is the last graph from the 27th:

Graph #1: 8Q subsets with 6Q Overlap, Data Sorted on X Values
You can see that the black line -- imagine the series of short black trendlines is one long line describing the pattern of dots -- the black line moves consistently rightward from lower to higher ratio values. That's because the data is sorted on those values.

When I put the data back in chronological order, the ratio values tend to get higher over time but there is some backtracking along the way. And a lot of the ratio values fall between 3½ and 5, so that we end up with a lot of dots and a lot of black trendline in that part of the graph:

Graph #2: 8Q subsets with 6Q Overlap, Data in Chronological Order
Quite a difference from the first graph. Quite a difference sorting makes.

Quite a difference, too, from my scatterplot of the unmolested data:

Graph #3: Unsorted Data, No Moving Averages, One Overall Trendline
No conclusion this time. Just pictures.

No comments: