Case Study Demographic data

The Eigenvector Research PCA/PLS toolbox (2004) is used to create a data set object named demoG. This file was created using CreateDemoG.m and can be loaded by typing load de-moG. A principal component analysis can be performed by typing pca. 

Let us autoscale the data, load it into the graphical user interface and selecting calc. Principal component calculation yield the results shown in Table 22.1. 

As can be seen, the first two principal components (PC s) capture already close to 75% of the variation in the data set. Select therefore two and select apply. 

If one would like to find out why these countries do not fit the model so well, one could look at the biplot, shown in Fig. 22.5. This plot is a combined scores and load plot. 

Singapore and Hong Kong do not fit the model so well, since they have a veiy high population per square kilometer and population per 1000 hectares agricultural land. India does not fit the model so well, since it has a high number of inhabitants per physician and high infant death. 

---

Case Study I. A supermarket chain with a large number of stores holds data on the shopping transactions and demographic profile of each customer s transactions in each store. Corporate management wants to use the customer databases to look for global and local shopping patterns. 

When modeling real data, an important first step is to carefully consider and specify the intended use of the model to be developed. With this in hand, an appropriate strategy and analysis plan can be crafted to ensure the appropriateness of the model (15). For the purposes of this chapter, a simulated data set is used to mimic a typical Phase 2 trial of a novel compound where an endpoint measurement is collected at each study visit. In this case, the endpoint of interest is the presence or absence of a particular adverse event. Various demographic variables are available for possible correlation with the endpoint in addition to a calculated measure of individual exposure to the drug. 

---