tobler.model.glm_pixel_adjusted¶
-
tobler.model.
glm_pixel_adjusted
(source_df=None, target_df=None, raster='nlcd_2011', raster_codes=None, variable=None, formula=None, likelihood='poisson', force_crs_match=True, **kwargs)[source]¶ Estimate interpolated values using raster data as input to a generalized linear model, then apply an adjustmnent factor based on pixel values.
Unlike the regular glm function, this version applies an experimental pixel-level adjustment subsequent to fitting the model. This has the benefit of making sure local control totals are respected, but can also induce unknown error. Use with caution.
- Parameters
- source_df
geopandas.GeoDataFrame
, required geodataframe containing source original data to be represented by another geometry
- target_df
geopandas.GeoDataFrame
, required geodataframe containing target boundaries that will be used to represent the source data
- raster
str
, required (default=”nlcd_2011”) path to raster file that will be used to input data to the regression model. i.e. a coefficients refer to the relationship between pixel counts and population counts. Defaults to 2011 NLCD
- raster_codes
list
, required (default =[21, 22, 23, 24]) list of integers that represent different types of raster cells. Defaults to [21, 22, 23, 24] whichare considered developed land types in the NLCD
- variable
str
, required name of the variable (column) to be modeled from the source_df
- formula
str
, optional patsy-style model formula
- likelihood
str
, {‘poisson’, ‘gaussian’} (default = “poisson”) the likelihood function used in the model
- source_df
- Returns
- interpolated
geopandas.GeoDataFrame
a new geopandas dataframe with boundaries from target_df and modeled attribute data from the source_df
- interpolated