The world's first energy market bagplot?

Trying not to try to notice what we never expected to see

More experiments with half-formed ideas and incomplete data. If you spot mistakes or can think of improvements or have context to share, please leave a comment.

The greatest value of a picture is when it forces us to notice what we never expected to see. John W. Tukey, Exploratory Data Analysis, 1977

We’ve made heavy use of box-plots in this series of short market notes. Plotting sets of box-plots over time has been useful to gain a sense of how Day-Ahead and Intra-Day VWAP price distributions have developed in time so far this year.

We need to keep looking at things from different perspectives. We can change perspective by changing what we look at or by changing how we look. We want to get the data to make faces at us.

What we’re really good at is comparing pictures in memory. You look at a face and you know immediately that’s Martin…

Last time we expanded our field of vision very slightly by adding imbalance prices to ID-price plots. This time, let’s try to find another way of looking at things. What do you think about this chart? I thought it was intriguing and immediately wanted to look at energy market data with it.

Source Link

This plot is called a bagplot. It’s a “bivariate generalization of the well known boxplot.” The inner dark coloured shape is called the “bag” and contains about 50% of all the points…

The light coloured loop plays the same role as the two whiskers in one dimension, so we could call [it] a “bag-and-bolster plot” to stress the analogy with the term “box-and-whiskers plot.” Like the univariate boxplot, the bagplot also visualizes several characteristics of the data: its location (the depth median), spread (the size of the bag), correlation (the orientation of the bag), skewness (the shape of the bag and the loop), and tails (the points near the boundary of the loop and the outliers).

https://wis.kuleuven.be/statdatascience/robust/papers/1999/rousseeuwrutstukey-bagplot-amstat-1999-color.pdf

Imagine a bagplot of ID1 VWAP vs. DA-Ahead data, say. Can you predict what you will see?