In UoA-eResearch/fisheR: Fisher's Information for timeseries data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Package info

This package for calculating Fisher's information follows methods from @cabezas_san_2010, @eason_evaluating_2012, and @ahmad_using_2016. Python code from @ahmad_using_2016 and MATLAB code from @cabezas_san_2010 has been adapted for modelling in R. For reproducibility purposes we use the 'sample_data' provided by @ahmad_using_2016 as an example which is the GDP per capita and total population of the USA from 1960 to 2013. For detailed descriptions on how to conduct Fisher's information see @cabezas_san_2010, @eason_evaluating_2012 and @ahmad_using_2016, we recommend reading these references first if unfamiliar with Fisher's Information.

Example run

The fisher function takes a dataframe as an input and an optional size of state (sos) dataframe. If a size of state dataframe is not provided one will be created following methods from @cabezas_san_2010, @eason_evaluating_2012 and @ahmad_using_2016.

The fisher function calculates: 1) the size of state; 2) Fisher's Information; and 3) Fisher's Information smoothed with a block mean of the desired window size. The final output is a dataframe where the last three columns are the Fisher's Information means, Fisher's Information smoothed and time-windows.

NOTE:

• Data structure for input dataframe must have the first column as time-steps and following columns containing the variables of interest

• NA values will be replaced by 0

• Take care with the header argument if reading in a csv

library(fisheR)
df = read.csv("../sample_data.csv", header=F)
sos = sost(df)
df_FI = fisher(df, sos = sos, display_plot = TRUE)

"To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment" - R. A. Fisher [@fisher_presidential_1938]

UoA-eResearch/fisheR documentation built on July 8, 2023, 5:06 p.m.