New folder inside scripts/ called sam_models/. The base level contains R scripts to organize datasets and their products as .Rda files. The subfolder mod1/ contains an R control script and a .JAGS model file.
Model description
Using chlorophyll A from station USGS 36, the dataset contains 280 observations of roughly bi-weekly observations spanning 2003-03-27 to 2019-04-25. The covariates are Q (from daily discharge at CCH, gapfilled by YOLO dayflow modeled output, cfs) and inundation (T/F). Q is scaled in the model.
The chlorophyll observations are log-transformed and given a normal likelihood. The mean mu is modeled as a linear combination of B coefficients, current day Q, and an antecedent variable of daily Q from one to eight days prior (to each chlorophyll observation). Each inundation status (1 = non-inundated, 2 = inundated) is given its own set of coefficients and antecedent weights.
Model runs and converges easily. The resulting posterior shows that for non-inundation periods, current Q is significantly positively correlated with chlorophyll, while antecedent Q is significantly negatively correlated with chlorophyll. The weight show that Q from one day prior has the strongest impact.
For inundation periods, the current model does not show significant coefficients or weights. Rather, the intercept has a large CI that is absorbing the variation. So, a different model form (or weight form?) might be more appropriate.
New folder inside
scripts/calledsam_models/. The base level contains R scripts to organize datasets and their products as .Rda files. The subfoldermod1/contains an R control script and a .JAGS model file.Model description