curve_fitting.do

// clear existing data and results (if any)
cls
clear

// load data from CSV
. import delimited "/Users/andreped/workspace/tumor-growth/data/fused_dataset_growth_analysis_remove-surgery_False_remove-missing_True_remove_multifocal_False.csv"

// convert string variable genders to categorical variable
encode patient, generate(patient_n)

gen log_volume = log(volume)

** Linear regression **
menl volume = {U0[patient_n]} + {xb:}, ///
	define(xb: i.gender_bin i.multifocality i.oedema i.t2 ///
	follow_up_months initial_volume initial_age)
	
estat ic

** Exponential regression **
menl log_volume = {U0[patient_n]} + {xb:}, ///
	define(xb: i.gender_bin i.multifocality i.oedema i.t2 ///
	follow_up_months initial_volume initial_age)

estat ic

// NOTE: Exp only better when T2 and Oedema is added because missing values in
//  t2 and oedema were removed!!! Hence, why "more" variance was explained!

** Linear radial growth regression **
menl log_volume = {U[patient_n]} + log(4 * c(pi) / 3) + 3 * log({xb:}), ///
	define(xb: i.gender_bin i.multifocality i.oedema i.t2 ///
	follow_up_months initial_volume initial_age)
	
estat ic

** Gompertzian regression **
menl log_volume = {U0[patient_n]} + log({a0}) - exp(-{xb:}), ///
	define(xb: i.gender_bin i.multifocality i.oedema i.t2 ///
	follow_up_months initial_volume initial_age)

estat ic