Are record temperatures indicative of a trend?

Not in climate science, no.

Many people make the assumption that past temperatures remain constant so a new high temperature implies an increase. With climate data, a new record can be attained because the earlier temperatures have fallen.

The other problem is that if you plot out the complete history of the climate over two sheets of graph paper then the thickness of that paper will cover a time span of about 10 million years. We only have thermometer readings for the last few hundred atoms-width.

yes their indicative of it being hot.

Idk

Yes, they certainly can be. For example for many years now we have had the number of record highs in the U.S. greatly exceed the number of record lows, this is what you would expect from a rising trend where we don't have a change in the detrended variance, as Yushchyshyn's example uses.

It will be very interesting to see how 2017 turns out after we have seen three record global high temperature years in a row, despite the end of last year having a La Nina condition.

I hope not, we just had SIXTY-FOUR DAYS IN A ROW of temps above 30oC, most of those being in the high 30s low 40s. Bugger that!

Using records or other types of ranked tests are non-parametric. They will certainly show a trend. What they will NOT do is help with future predictions.

This is rather simple to understand. A year can be called the "hottest" because it is 0.01 degrees hotter or 5 degrees hotter. While this is a LARGE difference, there is literally no difference for non-parametric analyses.

When you are trying to determine a trend just based upon how many places have a "record high" versus how many places have a "record low", it is a nearly useless analysis. The reason it is relatively useless is because you are dropping most all of your data. That would have to be the dumbest way I can come up with for finding a "trend".

Of course, every time I think I have heard the dumbest thing you can do, I am always shocked by the ingenuity of idiots.

You are the scientific type though, so you should understand the stupidity of attempting this method. Imagine you are running an experiment. You disregard 99.99% of the data and the ONLY data you actually collect are the extreme 0.01% and you only collect that information is a binary fashion. Sound like how you would proceed?

maybe