Working with the National Highway Traffic Safety Administration (NHTSA) database, the Fatality Accident Reporting System (FARS) and excel, I've figured out how to analyze the data to extract detailed records. For example, in 2007, there were these known accidents involving pedestrians and Prius: Column 1 Column 2 Column 3 Column 4 0 ST mm/dd/yyyy Vehicles Prius Model 1 CA 1/5/2007 1 2004 2 CA 1/24/2007 1 2006 3 FL 1/26/2007 1 2007 4 FL 9/12/2007 3 2006 5 TX 10/18/2007 2 2005 6 AZ 10/21/2007 2 2005 There were 5,432 records in the accident file giving a raw rate of 0.11% of all accidents with involving both a pedestrian and a Prius. In this rough report, there were three accidents involving more than one vehicle and pedestrian. A more rigorous analysis would take just the single vehicle pedestrian accidents. GET THE ACCIDENT DATA Start with NHTSA Encyclopedia Click on "FARS FTP" link at the top Select the directory for the year of interest Download the ZIP file of the data Open file into a drectory You should now see a list of files: ACCIDENT.DBF miacc.DBF midrvacc.DBF miper.DBF person.DBF vehicle.DBF vehnit.DBFDATA ANALYSIS If you DBase or Microsoft Access, these files should load fine and you can begin to analyze the data. However, I have a Macintosh and the built-in databases were unable to do anything with these files. In my case, a Macintosh with excel, which will "Open" any DBF file and each will load fully except for "person.DBF", which is too large to fit in one spreadsheet. But I only needed: ACCIDENT.DBF - all accidents for the year vehicle.DBF - all vehiclesThe first row will have the name of each field per column and each subsequent row has the values. Since it loads them as text and floating point numbers even though there are no floating point fields, I select the spreadsheet and format all numbers without any fractions. "ACCIDENT.DBF" This file lists each accident for the year including a column for "PED", pedestrians. So I simply deleted all non-pedestrian accidents and found 5,452 records. A stricter file would keep records with just one pedestrian and one vehicle, ignoring the multi-vehicle accidents that also involve a pedestrian. Now this file does not include vehicle information but it does include "ST_CASE" (state case number) that is a universal index in all of the files. "vehicle.DBF" This file lists all vehicles involved in an accident and has a column for "ST_CASE" but in the wrong location. I make a copy of that column and do an 'insert paste' into the first column of the "vehicle.DBF" file. This will be the key needed to lookup each vehicle. But you will also need a number for each column. Going to the bottom of the spreadsheet, I add another row that is a counter for each column with the "A" column being "1." Adding Vehicle Information To ACCIDENT.DBF Columns Select a column in "ACCIDENT.DBF" and insert the function "vlookup". This function needs: ST_CASE cell - using an absolute reference for the column and relative row, identify the cell with the state case number. vehicle.DBF table - using just the data fields, not the title row nor column count row, identify the whole spreadsheet data as a table. On a Mac, I click on the upper left data cell and then SHIFT-click on the lower right. vehicle field "MAKE_MOD" - this field includes both the manufacturer and the last thee digits are the model number. The number 49046 is the Toyota Prius. Add as many additional vehicle fields as desired in additional columns. UNDERSTANDING THE DATA FIELDS Using Google, find a copy of "Analytic Reference Guide 1975 to 2006" (FARS06.pdf) for a text description of the data fields. WHAT TO DO NEXT? It would be great if we could find a gas-only vehicle with known sales figures and a similar sales pattern to the Prius to compare and contrast. For example, the Scion and Prius might give some insights if one or the other shows an abnormally high rate of pedestrian accidents versus their sales numbers. Anyone interested in helping with say the Scion sales figures? Comments? Questions? Suggestions? Bob Wilson ps. I wrote a quick "c" hack to pass through the "person.DBF" file and generate a comma delimited, text file. The tricky part was dealing with the binary 'endian' but once solved, it works well enough. This is not a 'general purpose' DBase file program, no run-time arguments, everything is hard-coded as needed to get the data. But if you have an interest, send me a PM and I'll forward the source.