On the use of regression models in the study of road accidents

THE EXPECTED NUMBER OF ACCIDENTS A CERTAIN DAY IS ASSUMED TO DEPEND ON SOME REGRESSION VARIABLES. THE NUMBER OF ACCIDENTS IS ASSUMED TO BE NORMALLY DISTRIBUTED, THE VARIANCE BEING CONSIDERED FOR ALL THE DAYS CONSIDERED. NUMBER OF ACCIDENTS ON DIFFERENT DAYS ARE FURTHER ASSUMED TO BE INDEPENDENT. THE REGRESSION PLANES CORRESPONDING TO DIFFERENT PERIODS CAN THEN BE COMPARED BY THE HELP OF SIMULTANEOUS CONFIDENCE REGIONS. THE ASSUMPTIONS ARE DISCUSSED AND ELUCIDATED BY DATA AND CONFIDENCE REGIONS ARE ILLUSTRATED. COMPARISONS ARE ALSO MADE WITH THE RESULTS OBTAINED WHEN THE NUMBER OF ACCIDENTS IS ENSTEAD ASSUMED TO BE POISSON DISTRIBUTED. /AUTHOR/