Massive Overhaul #17
Massive Overhaul #17
Conversation
Starting a notebook for uncertainty analysis. For now, we are only to the point that we have a plot of correlation vs. offset for one subset of one profile.
The velocity calculation was being done for all beams in a loop within the calculate function. Trying to clean this up.
There are way too many separate arrays. I am moving them all into a dictionary for now. It might be better to restructure this as a class eventually, but it might be too big a step for the group right now.
Separating processing functions and limiting the number of variables that we are using.
Probably at the stage that I can have Grace look at it.
|
Looks great, Ben. Looks like you also added a step to interpolate places where there are nans? I think we don't want to interpolate nans linearly, b/c lines have unexpected behavior in the correlation functions - ex., a flat line correlated with a flat line might be highly correlated, despite there being no signal. We'll have to think more about how to fill in gaps / if we need to. |
|
oh, also, we need to retain options for the 'filt' function |
|
Ok. I will get on both of these points. Thanks for looking at it. I am going to try to do this today and then I can start making some figures. |
|
So, the options for the filt function are still there, they just default to the values that we had been using a lot. If you want to change them though, they still exist. In regard to the nan interpolation, I just copied that from the loop. I think it was being done in previous versions, maybe that was unintentional though. Anyway, it is not hardcoded into the velocity function, just an additional function. Maybe I will add a warning to that funciton. |
Bringing in the reorganization changes that were made.
We were keeping track of a bunch of different variables like 'h_full', etc. I just streamlined this but added a flags field so that we can tell what processing steps have been done.
I wanted to remove the dependency on the MOA file.
I think that the conversion was ignoring one of the velocity components. I tested this updated conversion for all combinations of +/- vx and vy terms.
…correlation function; update tutorial notebook to reflect
In converting from Measures Velocity (x/y) to along/across we were originally using only the first angle. However, the angle changes over the length of the track, so calculating a full array is preferred.
re-arrange correlation steps slightly; add function to plot a single …
Get Measures Velocities
I wanted to break out some of the functions so that it is more clear where people can plug in. I also moved all the variables into a dictionary so that we are not importing and exporting so much from each function. We talked about using a Pandas dataframe... I think the best way to do this if we wanted to be more 'proper' would be to make a data class and put all the variables as well as most of the functions inside of that (low priority for now though).
I feel like I am getting closer to the point that hopefully others can hop in and make some figures.
Grace - Before we merge this branch, can you get onto devel and make sure that it works for you?
Edit: Waiting on #19 before merging this.