My city has an open data section listing crimes and accidents amongst other things. They include approximate time, type of crime, etc. The way they formatted the location Looks something like this "42XX street, city, state, zip", how would I geocode an address that doesn't have a a complete street number? If at all possible. I'm looking to incorporate census tract data as well so I need good spatial accuracy.
Lets say that you want to geocode the following address:
42xx Ashland Ave Baltimore MD 21205
One method would be to add the previously mentioned "50" to get 4250. Geocoding this will get you pretty close.
Another method, since you know that it will definitely be between 4200 and 4299, would be to attempt validating each address. Yes, all 100 of them. The outcome of this method would be the following:
4200 ASHLAND AVE BALTIMORE MD 21205-2920
4208 ASHLAND AVE BALTIMORE MD 21205-2920
4209 ASHLAND AVE BALTIMORE MD 21205-2922
Those are currently the only valid address on Ashland Ave in Baltimore between 4200 and 4299. Running that through a geocoder could give you an exact lat/lon pair for the addresses.
So, it just depends on what you would like to end up with. If just an approximate lat/lon will work, stick with the first method. If you want to be more precise by determining the potential valid addresses, go with method two.
A quick google search for "address validation" will give you a number of address validation services. I work for one called SmartyStreets.