Stack Exchange network consists of 183 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.
My approach in situations like this is to estimate bounds - for example, how many geographic units do you have? Using those as the unit of analysis would give you a lower bound, while ignoring the spatial auto-correlation would give you an upper bound. This is a very interesting problem, and I look forward to seeing what suggestions you get.
This is somewhat of a subjective question, I think, but I would say that you are only really missing one important relationship - the mediator (intermediate). Basically all other relationships can be explained by combinations of these. I'm not sure if it's possible to use latex packages in a comment (specifically, tikz), but if anyone knows how to do that, I will add an answer with the DAGs.
I'm not sure I understand why the averages are needed at all. If the goal is to reach a target, I would suggest that you look annually at distance from the target or the difference from baseline. But, perhaps the targets are defined on a multi-year basis or the underlying process takes a while to change? In which case, I think that a weighted average might be preferable since it will be more responsive to any programmatic changes you have implemented in the past year. The mean of T5 and WTA is just a weighted average where T5 has a weight of 6 instead of 5, which seems unnecessary.
If you are comparing the outcomes, you should be using the same methods. But, when you say 'baseline', presumably you mean that you will compare these to some rates in future years after one or more intervention, rather than comparing between the outcomes. In that case, the bigger issue is how you will do the comparison when there are obvious time trends in your data. You will probably want to use a method that takes this into account, rather than averaging the baseline measures. If you can give more detail on the comparison(s) you want to make, we can suggest appropriate methods.
