All black boxes leave enough breadcrumbs to follow to figure out a great deal about what makes them work. Or in the case of CSI what makes it suspect.
CSI did not provide any information about how it was derived, so the black box couldn’t be cracked open to look inside. But there were enough breadcrumbs to follow conclude that the result produced by the CSI black box is so flawed that the input to the black box is necessarily flawed as well. Garbage in – Garbage Out.
To do a CSI type predictive analysis of crime for specific locations in the Caribbean using so many different measures (or variables), the methods would involve something called multiple regression analysis. That only means CSI would be taking many measures of crime and socio demographic measures (like unemployment, literacy, economic activity etc) and trying to see if they will give a probability that a location will have a lot or a little bit of crime.
To calculate the probabilities well, good data has to be available for the location under study. Unfortunately, for the locations CSI claims to calculate indices, insufficient, unsystematically collected data are all there is. Literature searches and data searches can not locate data that would meet professional standards (below) that could render a CSI in its present form. CSI has not responded to any request to identify even generally what data was used for each of the indices published. Which strongly suggests there is something to hide. Something like flawed data?
The point is that it is likely any data used in the inner workings of the CSI Black Box was either made up, too general, not current, over processed or shamelessly massaged until it fit preconceived notions.
The data cocktail consumed by the CSI black box has made it dangerously impaired.
In a phrase – garbage in – garbage out.
If you are a statistical nerd you may want to follow the source links below to see the standard required of data that would be used in a CSI type analysis for a widely used statistical package. Because the present socio demographic and related crime data available for the Caribbean can not meet those standards, we must conclude CSI is likely pseudoscience and its data input was manipulated toward a biased conclusion.