Daily Archives: November 5, 2018

Multivariate Data

In this week’s blog assignment, I examined the relationship between calories, potassium and fibre using three different tools. They include a scatterplot matrix, coplot and spinning scatterplot.

Figure 1: Scatterplot matrix of potassium, calories and fibre with loess curve spanned at 0.8.

Figure 2: Coplot of fibre against potassium, conditioned on calories.

Figures 3 – 7: Snapshots of spinning scatterplot
  

Using all three tools, we see that there is a positive, moderate and linear relationship potassium and calories. Generally, an increase in potassium includes an increase in calories as well. From the scatterplot matrix, we see that as potassium reaches 400 grams, the amount of calories gradually levels out. Clearly, we can see that there are outliers and a few possible cases of high leverage.

On the other hand, the relationship between potassium and fibre is strong, positive and linear. That is, small amounts of potassium tend to have small amounts of fibre. Likewise, large amounts of potassium generally have larger amounts of fibre as well. Again, there are three data points that can be considered influential points. In other words, it seems that these three points have potassium and fibre amounts that are far from the other points.

Lastly, the relationship between calories and fibre is weak, positive and linear. A small amount of calories tends to have less fibre, vice versa. From the plots, we can see that there are a couple of points that move away from the loess curve. Some of these points are outliers, meaning that their amount of fibre deviate from the pattern of other cereal brands. Furthermore, some of these points are high leverage cases, meaning that their amount of calories stray away from the pattern of other cereal brands.

Two cereal brands that deviate from the general pattern of the scatterplot are Grape Nuts and 100% Bran. Grape Nuts is a “special” brand because for the amount of potassium that it has (360 grams), the cereal has a really high level of calories (450 cal). Most brand with 360 grams of potassium only have between 150 to 300 calories. 100% Bran is also considered a “special” case because it has a very high amount of fiber (30 g) for having only 212 calories. Other cereal brands around 200 calories have 10 or less grams of fiber. However, 100% Bran’s fiber amount is considered very large compared to the patterns of other cereal brands.