My understanding is that write back policies modifies only in cache and sets a "dirty" bit on the block and write through modifies both in cache and memory. When a replacement is done, if it's using a write back policy, it will check if the block is dirty to replace it in memory.
My question is, which of these policies generates more traffic? Is there any other parameters that will affect the increment and decrement of the traffic on a given policy?
Memory traffic depends on the application, both schemes have their merits. But just the write policy is not sufficient to analyze the traffic.
You also need to consider what sort of write miss policy you are going to implement. (Write allocate vs no write allocate). The former, on a cache miss writes just the modified entry to an empty cache line and keep a mask to remember what words are real and what are garbage. The latter, on the other hand treats write misses similar to read misses and bring the entire cache line from memory and modify the required word. Both schemes have their merits and usually each has a preferred write policy (write back vs write through). Here is a good article by Norman P. Jouppi on writre miss policies. http://dl.acm.org/citation.cfm?id=165154