Full static network based on DAG output

rbartelme commented 4 years ago

Continuing with results from issue #15:

[x] Look for spurious associations/derived data links
[x] drop derived data links from the dataset
[x] compile in report with first structure learning algorithm (continued in issue #14)

rbartelme commented 4 years ago

Moved to Sprint 35, and opened issue #15, instead of issue #12 output (which is cumbersome). I'll use issue #15's output to look at the network.

rbartelme commented 4 years ago

Fixed task list to reflect the shift in network dataset from issue #15.

rbartelme commented 4 years ago

Derived data links are obvious and will be blacklisted for the full run (i.e. min/max/mean by day)

NA's led to spurious network from the tall to wide format conversion.

rbartelme commented 4 years ago

First Static Network Attempt Report:

by Ryan Bartelme, PhD - June 4th, 2020

First attempts at automated learning were with mixed results.
It's probably best to think about the biology and physics of the system, and hardcode a graph structure. Then it will be easier to fit the raw data to the graph structure and see what the probabilities are at each node.

Cultivars are not moving, they are statically linked to their location
Canopy height is linked to vapor pressure deficit (Poor, 2019)
min/max/daily mean are linked...but maybe choosing the mean value would be best?
The data are ordinal, that is important to specify - maybe it's important to make the time connection now?

rbartelme commented 4 years ago

Reports on network efforts will reside here. The reports will be filed in the github issue as it is closed, and copied to the linked directory.