I did some searching on that topic, but I couldn't find much. I'm doing intense training and will be switching between gaining and cutting. Last month, my weight loss was right on what was expected based on the calorie deficit. That is after accounting for going on creatine and also considering glycogen depletion. I took a baseline weight measurement at the beginning of the month before creatine and then came off creatine specifically to check weight. I also made sure I wasn't especially loaded or depleted of glycogen (huge water weight swings possible) relative to the first of the month. Overall it seems accurate so far, but I'll really need to watch it for several months to be sure.

As for my own trying to figure it out before getting a few more months of data in, I have a few thoughts. Looking at several estimated methods (Endomondo, HR formula, etc.), I see a range for an hour of weightlifting burning about 300 to 1000 calories. That's just useless compared to using any sort of device (HRM or BM). It does feel like BodyMedia is accurate, at the least, it's certainly consistent. I can see the log is ~50% higher for the more intense sessions. If anything, I suspect it could be measuring a bit low for me since the motion sensing aspect is often not in play as much during weight training as compared to jogging (especially when keeping arms still while killing it on the legs), and I just don't sweat a lot even when my HR is 150-180. Therefore with the galvanic skin response and accelerometer in question, that just leaves heat and heat flux as the primary measures. Of course, I don't know how sophisticated their algorithm is to account for those things though, so I may find it is dead on.

I think that even if I find that it is off by some amount for weight training, it will likely be by an average percentage that can be factored in. I should have a better idea in a few more months.