Comments: | This pair of CMF and CRF equations was obtained from Model #7 (see pg. 57).The researchers removed two outliers prior to modeling (p. 55). This fact must be kept in mind when computing sample size, average ADT, etc.Only 1 week of field data was collected to estimate the traffic volumes and 85th percentile speeds, and no traffic volume conversion factors were used to adjust for seasonal variations in traffic. On the other hand, 10 years of crash data were used. Thus, there was a major disparity in the accuracy of the response variable relative to the predictor variables. The functional classification of the sample used was never specified. The researchers began with crash data from all functional classifications (p. 48), but they screened out the all but the rural crashes. The tables in the report describing the sites (e.g Table 5.1 on p. 49) do not state functional classification either. |