@map7 You would be comparing two different 24 hour windows, which is not scientific. When testing for a specific outcome, you must eliminate all other probable causes. In the case of BME680 I might set up another software i2c bus on two other GPIO pins, connect both sensors via jumper jerky at the same time, sit them side-by-side, and run them simultaneously, comparing the results over time. In this case I might expect some variation in temperature due to thermal coupling through the wires, but not much.
@badboybubby it’s probable your Rainbow HAT is just thermocoupled to your Pi via all the ground connections in the ribbon cable. Since the ground plane runs through the whole board, it’s not difficult for it to get and stay warm. In cases like this, you must compensate the temperature reading in software.
Initial State have a good article about temperature compensation here: http://blog.initialstate.com/tutorial-review-enviro-phat-for-raspberry-pi/