-If we wish to plot total number of drivers involved in crashes by state, I believe we CAN simply sum the weights in each state for the "drivers only" data set -- even though the full accident weight is assigned to each row (i.e. each driver). This is because an accident with 5 drivers should count the weight of that accident 5 times to get the appropriate weighted sum of drivers in the population. Example: if an accident with weight 22 has 5 drivers, that means there are estimated to be 22 such crashes that involve 5 drivers. Therefore, that crash represents 110 total crash-drivers in the population.
-For this same reason, we do not need to drop the accident with 2 drowsy drivers when doing this plotting.
Patrick wrote:
Notes:
-If we wish to plot total number of drivers involved in crashes by state, I believe we CAN simply sum the weights in each state for the "drivers only" data set -- even though the full accident weight is assigned to each row (i.e. each driver). This is because an accident with 5 drivers should count the weight of that accident 5 times to get the appropriate weighted sum of drivers in the population. Example: if an accident with weight 22 has 5 drivers, that means there are estimated to be 22 such crashes that involve 5 drivers. Therefore, that crash represents 110 total crash-drivers in the population.
-For this same reason, we do not need to drop the accident with 2 drowsy drivers when doing this plotting.