MTBF calculations and statistics in many circumstances can be a waste of your time. Even more often, they can be misleading. There is a great example of MTBF (failure rate) being a waste of time in the article "PLC Controller Failure Rate". This MTBF article explores in great detail why MTBF doesn't work with PLCs, and via the analysis, shows you a much more reliable way to approach failure calculations in this particular case example.
Side Note: PLC stands for Programmable Logic Controller. PLCs control almost all machinery in all industries, in every country. They also control all infrastructure around the world. Recently in the last 10 years, they have evolved to PACs (Programmable Automation Controllers) which instead of being designed to be lectrician user-friendly like the PLC, the PAC is designed to cater to computer programmers and IT personnel. So now not only should all electrical engineers be well versed in PLCs, so should Computing Engineers and IT. See http://bin95.com/PLC-PAC-Difference.htm if you want to start learning now.
As reviewed in the PLC Controller Failure Rate article above, "Time" is the most important consideration when deciding if MTBF will work for your application. If the item has a constant use, based on hours of operation like electronics with a steady current, the MTBF is a good measurement. (as far as those stats being available and you can verify your equipment's environment is similar to the vendor's environment when they tested for MTBF) But if you equipment has mechanical attributes with sporadic use, other failure analysis methods should be used. In the above article, one aspect examined was relays (electro-mechanical) which number of activations vary greatly depending on PLC program.
I highly recommend you read the article "PLC Controller Failure Rate", and save yourself some time and money instead of using MTBF.