Factors not considered with camera numbers

In statistics; the person doing the statistical assessment, if they are not careful (or intentional), determine statistical significance. This came to mind in the March 14 article (“Speedcheck”) about the two-year anniversary of the traffic cameras.

Many numbers were thrown accidents, also in selection of various reporting years. It appeared that the points used were selected for significance and not truly representative of the cameras'' effects.

A new car will get 500 miles on a tank of gas, great, right? If it is a 100-gallon tank, not so great. The same is true of the numbers in the article without normalization. You will have more accidents on 1 million vehicle miles than 100 just by sheer opportunity. Normalization allows for reasonable comparisons.

The second issue is variation. Is the noted change greater that the natural variation of the system? Just a cursory look at the data shows significant variation year to year. This could be because of weather, traffic patterns, driver behavior, volume of traffic and many others. Without this calculation of variance, the city's statements cannot be supported.

A simple analysis of the data over time would show conclusively what the various factors are what affect the accident rate. Normalization of the data to miles traveled or traffic volume would remove some of the above factors as contributors.

Richard Percifield

Norway