5. Agents

A series of tools to perform data manipulation and adaption.

5.1. Capping

Cap data to certain values.

5.1.1. CapAgent

Class for capping. The class can be instantiated without specific parameters. Arguments can be later specified with self.fit() method.

resourcegeo.transformation.capping.CapAgent.fit(self, greater_than=None, less_than=None, cap_to=None, subset_col=None, subset_code=None)

Define required capping parameters.

Parameters:

  • greater_than (float) – threshold value

  • less_than (float) – threshold value

  • cap_to (float) – capping value

  • subset_col (str) – column name in pd.DataFrame

  • subset_code (int) – categorical code

Returns:

Class object with fitting parameters.

Return type:

self

resourcegeo.transformation.capping.CapAgent.fit_apply(self, data, var, suffix='_cap')

Cap values of values in a dataframe column. It requires the fitting parameters previously specified with self.fit(). It does not require and does not change initialized weights.

Parameters:

  • data (pd.DataFrame) – df with values

  • var (str) – column name for values

Returns:

None

resourcegeo.transformation.capping.CapAgent.sensitivity(self, values=None, limit_value=None, weights=None, wt_col=None, chosen_value=None, show_progress=False)

Capping sensitivity analysis over a range of capping values.

Parameters:

  • values (int or list[float], np.array(float)) – If int, it is the number of capping values for the sensitivity. If list or np.array, it is the values to use for sensitivity

  • limit_value (float) – If None and capping above a threshold, the limit value is the maximum value of uncapped data. If None and capping below a threshold, the limit value is minimum value of uncapped data. The limit_value can be also specified by the user.

  • weights (np.array) – weights for calculating statistics. It must be same length than the data.

  • wt_col (str) – column name in self.data for the weights.

  • chosen_value (float) – If value is not in values, it is added.

Returns:

None

resourcegeo.transformation.capping.CapAgent.sensitivity_plot(self, line_at=None, title=None, lw=0.5, figsize=None, xscale='log', sinch=(8, 6), fontsize=8, ms=1, nbins=None, flname=None)

Plot a set of four plots for capping sensitivity analysis. Plot a user defined vertical capping line in the graph and show the closest response to that value. If weights were used in sensitivity(), then the statistics in the histogram and the probability plot are weighted.

Parameters:

  • line_at (float) – Value to plot a vertical line in the histogram and cdf subplots. If summary of the response is shown in the graph, it corresponds to the closest capping value to line_at

  • title (str,optional) – title of the graph

  • lw (float) – width of the vertical line plotted at line_at value

  • figsize (tuple(float,float)) – figure size

  • xscale (str) – x-scale ‘log’ or ‘linear’ for the histogram and cdf

  • fontsize (float) – font size

  • ms (float) – marker size for third and fourth subplot

  • nbins (int) – number of bins for the histogram

Examples:

import resourcegeo as rs
df = rs.BaseData('assay_geo').data
capped = rs.CapAgent().fit(greater_than=1,cap_to=1,subset_col='rock',subset_code='MXPRI')
capped.fit_apply(df,'CUpc')
capped.sensitivity(values=200, wt_col='ai')
capped.sensitivity_plot(line_at=1)
_images/agents-1.png
resourcegeo.transformation.capping.CapAgent.unistats(self, weights=None, wt_col=None)

Obtain summary statistics of the capped values

Parameters:

  • weights (np.array) – weights for calculating statistics. It must be same length than the data.

  • wt_col (str) – column name in self.data for the weights.