Why not starting with some example empirical measurements?
Lets try with a 802.11b (rather old and relatively slow) network. My effective best throughtput in data transfer uses to be 1MB/seg for this network:
Case 1: Best scenario. Only one file.
Number of files: 1
Total transferred size: 37.060.608 bytes
Average size/file: 37.060.608 bytes
#Data captured: 41.478
Case 2: Average scenario. A few files.
Number of files: 11
Total transferred size: 6.150.538 bytes
Average size/file: ~500.000 byte
#Data captured: 10.772
Case 3: Worst scenario. Lots of small (1 byte size) files.
Number of files: 10.000
Total transferred size: 10.000 bytes
Average size/file: 1 byte
#Data captured: 134.270
So, results vary widely from nearly 1 MB/data to an astonishing less than 1 byte/data (LOL).
If I am not wrong, we can conclude that, without more info about the data being transferred, and knowing only the
#Data field, there is no practical way to know the correlation between
bytes on the wireless traffic.
Maybe this conclussion could be true for cable networks too.