Different # chems across samples

[limnoliver]

Pulls from NWIS show a different number of chemicals that were measured per sample. This is partially because neonics and glyphosate not measured in every sample, but we know when to expect this. Other possibilities include possible lab issues where single chemicals may have been dropped.

Some issues with there being different n chems across samples: not comparing apples to apples across samples - so when summing something like n chems detected or EAR for a sample, you're not comparing the exact same thing.

Steve and I talked about a variety of approaches to both figure out why there are different numbers of chemicals across samples, and then to explore what the consequence of that is.

Why

1. Count n chems with and without neonic schedule included, or compare n chems to n chems expected based on tracking info. Or count n chems by each group (pesticide schedule, neonic schedule, glyphosate, etc).
2. Start by looking at pesticide schedule, find cases where schedule is missing one or two compounds, and cross reference the NWQL sample status to see if there are comments about that compound/sample combo.

Consequences

1. First generate the entire chemical (y axis) x sample (x axis) space and mark where we have missing data. This will provide a stat/visual for where and how often we are missing data across samples. Additionally, could id cells where chemicals were actually above DL. We may be missing some chemicals for some samples, but if they're rarely detected, it likely doesn't matter. Later, we could also use some approaches to gap fill. E.g., some machine learning approach to use relationships between all chemicals to fill in missing data.
2. Use neonics as test case since there are many cases where we have them and don't have them at the same site or date. How often do neonics contribute to EAR values? How often are they above the benchmarks, or even detected?