A statistical error, a Sample of Convenience is when data is used that is "easily acquired" instead of "equally acquired." Network Applications is able to detect which browser is used by those who went to the pages Network Applications chose to monitor.
Here's an edge case. Let's say that all users of X-OS are stupid people who love celebrities. Someone observing the X-Kardashian web site might note that lots more people use X-OS. This is a sample of convenience. It doesn't take into account the source of the traffic.
Another example: go to an upscale shopping mall. Watch phones. The convenience sample is NOT of "phones" but rather "phones used by people in public who shop at an upscale mall." Perhaps I, a discriminating polite person, will not pull my phone out in a mall for no reason. Perhaps I will pull out my Nokia N900 and not the Android phone. In either case, there are other factors, but it's "convenient" to do NOTHING to remove the bias.
I sure hope Java ME and IOS and Android continue to dominate. I also hope SURVEY TAKERS who do a piss-poor job of removing bias GO OUT OF BUSINESS.