this question doesnt make any sense without the frequency, you cant just look at timing data
but *IF* you're somehow comparing otherwise identical memory, same size, same frequency but with marginally different timings, the only item you need to worry about is the FIRST number, which the CAS Latency (CL) value.
in this case 14 is better than 15 is better than 16.. but *ONLY* if the frequency is the same.
but really these spec are only useful for matching memory, not as sales/marketing hype.