This report is the first of a series in which data sets gathered by others are revisited. As such, it suffers several weaknesses. Firstly, it is somewhat bulky in volume and didactic in tone because of the need to develop methodology and convincingly communicate it to the reader. Secondly, as it is written, certain portions already appear inadequate.. Thus, for example, we now realize that (the index of effectiveness depends on labda - the expected number of accidents). In consequence, one should revise the likelihood function L (theta) to make it depend on labda. It is perhaps the nature of research that the opening of a door hitherto locked, reveals several other doors which require opening. Even so, much has been accomplished. For one, we now have good estimates of the safety effect that the conversion from two-way to four-way-stop control had on several accident types at 49 intersections in San Francisco. Perhaps more importantly, we have the rudiments of a much improved methodology that promises to yield more precise estimates and more detailed insights. Finally, we cast our findings into a form which now invites comparison with findings in other cities and, perhaps, aggregation into a coherent body of knowledge. (Author/publisher)
Abstract