prevalence_predictor_mgcv: Function to predict prevalence excedance probabilities at...

In disarm-platform/DiSARM: A collection of spatial analysis functions related to the DiSARM project at UCSF

Description

Function to predict prevalence excedance probabilities at spatial locations. Currently only support binomial data.

Usage

prevalence_predictor_mgcv(point_data, layer_names = NULL, v = 10,
  exceedance_threshold, batch_size = NULL, uncertainty_fieldname,
  additional_covariates = NULL,
  covariate_extractor_url = "https://faas.srv.disarm.io/function/fn-covariate-extractor",
  seed = 1981)

Arguments

point_data	Required. An sf object of points containing at least 'n_trials', 'n_positive' fields
layer_names	Optional names of column corresponding covariates to use. Choose from 'Layer names' as outlined [here](https://github.com/disarm-platform/fn-covariate-extractor/blob/master/SPECS.md). If none provided then spatial only model assumed.
v	The number of folds to use in the machine learning step. Defaults to 10.
exceedance_threshold	Required. The prevalence threshold for which exceedance probabilities are required
batch_size	The number of adaptively selected locations required. Defaults to NULL.
uncertainty_fieldname	If 'batch_size' is specified (>0), adaptive sampling is performed. To sample in order to minimize classification uncertainty choose 'exceedance_probability'. To sample in order to minimize prediction error choose 'prevalence_bci_width'.
additional_covariates	Optional vector of column names of 'point_data' referencing additional covariates to include in the model. Defulats to NULL
covariate_extractor_url	The function currently makes use of the temporary DiSARM API function 'fn-covariate-extractor' to extract values of 'layer_names' at locations specified in 'point_data'. If this algorithm is hosted somewhere other than the DiSARM API, include the URL here.

disarm-platform/DiSARM documentation built on March 4, 2020, 3:49 p.m.