There is a difference

So having finally extracted some data from the Owl monitor (more on this another day), I'm in a position to compare the system output from before and after the scaffolding was removed.

The results are pretty clear. The chart below shows the comparative data for the cumulative generation for the 12 days with and without the scaffolding.

It's fairly clear here that there is a difference, the scaffolding was reducing output by about 50%, which corresponds with the level of drop I was expecting to see from my research.

To be really nerdy, I also put the individual days of data through a t-test, based on a single-tail test of un-paired data. The result again was clear, with p<0.005. I.e statistically there is 0.5% chance that the data difference was accidental and a 99.5% chance it was a real difference. That's well beyond normal statistical tests and good enough for me.