Biogeme utilities
Some useful additional pieces of software.
This utility reads a model description file written for the Bison
version of Biogeme (.mod) and transforms it into a model
description file for the Python version of Biogem (.py).
Syntax: mod2py model transforms the
file mymodel.mod into mymodel.py.
Example, .mod file:
[Choice]
CHOICE
[Beta]
// Name Value LowerBound UpperBound status (0=variable, 1=fixed)
ASC_CAR 0 -10 10 0
ASC_TRAIN 0 -10 10 0
ASC_SM 0 -10 10 1
B_TIME 0 -10 10 0
B_COST 0 -10 10 0
[Utilities]
// Id Name Avail linear-in-parameter expression (beta1*x1 + beta2*x2 + ... )
1 A1_TRAIN TRAIN_AV_SP ASC_TRAIN * one
+ B_TIME * TRAIN_TT_SCALED
+ B_COST * TRAIN_COST_SCALED
2 A2_SM SM_AV ASC_SM * one
+ B_TIME * SM_TT_SCALED
+ B_COST * SM_COST_SCALED
3 A3_Car CAR_AV_SP ASC_CAR * one
+ B_TIME * CAR_TT_SCALED
+ B_COST * CAR_CO_SCALED
[Expressions]
one = 1
CAR_AV_SP = CAR_AV * ( SP != 0 )
TRAIN_AV_SP = TRAIN_AV * ( SP != 0 )
SM_COST = SM_CO * ( GA == 0 )
TRAIN_COST = TRAIN_CO * ( GA == 0 )
TRAIN_TT_SCALED = TRAIN_TT / 100.0
TRAIN_COST_SCALED = TRAIN_COST / 100
SM_TT_SCALED = SM_TT / 100.0
SM_COST_SCALED = SM_COST / 100
CAR_TT_SCALED = CAR_TT / 100
CAR_CO_SCALED = CAR_CO / 100
[Exclude]
(( PURPOSE != 1 ) * ( PURPOSE != 3 ) + ( CHOICE == 0 ))
[Model]
$MNL
Generated .py file:
# This file has automatically been generated.
# Tue Apr 21 09:36:59 2015
# Michel Bierlaire, EPFL
# biogeme 2.4 [Lun 19 jan 2015 18:40:42 CET]
# Michel Bierlaire, EPFL
#####################################################
# This file complies with the syntax of pythonbiogeme
# In general, it may require to be edited by hand before being operational
# It is meant to help users translating their models from the previous version of biogeme to the python version.
#####################################################
from biogeme import *
from headers import *
from loglikelihood import *
from statistics import *
# [Choice]
__chosenAlternative = CHOICE
# [Weight]
# NONE
#[Beta]
#Parameters to be estimated
# Arguments:
# 1 Name for report. Typically, the same as the variable
# 2 Starting value
# 3 Lower bound
# 4 Upper bound
# 5 0: estimate the parameter, 1: keep it fixed
ASC_CAR = Beta('ASC_CAR',0,-10,10,0)
ASC_SM = Beta('ASC_SM',0,-10,10,1)
ASC_TRAIN = Beta('ASC_TRAIN',0,-10,10,0)
B_COST = Beta('B_COST',0,-10,10,0)
B_TIME = Beta('B_TIME',0,-10,10,0)
# [Expressions]
# Define here arithmetic expressions for name that are not directly
# available from the data
one = DefineVariable('one',1)
CAR_AV_SP = DefineVariable('CAR_AV_SP', CAR_AV * ( SP != 0 ))
TRAIN_AV_SP = DefineVariable('TRAIN_AV_SP', TRAIN_AV * ( SP != 0 ))
SM_COST = DefineVariable('SM_COST', SM_CO * ( GA == 0 ))
TRAIN_COST = DefineVariable('TRAIN_COST', TRAIN_CO * ( GA == 0 ))
TRAIN_TT_SCALED = DefineVariable('TRAIN_TT_SCALED', TRAIN_TT / 100 )
TRAIN_COST_SCALED = DefineVariable('TRAIN_COST_SCALED', TRAIN_COST / 100 )
SM_TT_SCALED = DefineVariable('SM_TT_SCALED', SM_TT / 100 )
SM_COST_SCALED = DefineVariable('SM_COST_SCALED', SM_COST / 100 )
CAR_TT_SCALED = DefineVariable('CAR_TT_SCALED', CAR_TT / 100 )
CAR_CO_SCALED = DefineVariable('CAR_CO_SCALED', CAR_CO / 100 )
#[Group]
#[Utilities]
__A1_TRAIN = ASC_TRAIN * one + B_TIME * TRAIN_TT_SCALED + B_COST * TRAIN_COST_SCALED
__A2_SM = ASC_SM * one + B_TIME * SM_TT_SCALED + B_COST * SM_COST_SCALED
__A3_Car = ASC_CAR * one + B_TIME * CAR_TT_SCALED + B_COST * CAR_CO_SCALED
__V = {1: __A1_TRAIN,2: __A2_SM,3: __A3_Car}
__av = {1: TRAIN_AV_SP,2: SM_AV,3: CAR_AV_SP}
#[Draws]
BIOGEME_OBJECT.PARAMETERS['NbrOfDraws'] = "150"
#[Exclude]
BIOGEME_OBJECT.EXCLUDE = ( ( PURPOSE != 1 ) * ( PURPOSE != 3 ) ) + ( CHOICE == 0 )
#[Model]
# MNL // Logit Model
# The choice model is a logit, with availability conditions
prob = bioLogit(__V,__av,__chosenAlternative)
__l = log(prob)
# Defines an itertor on the data
rowIterator('obsIter')
# Define the likelihood function for the estimation
BIOGEME_OBJECT.ESTIMATE = Sum(__l,'obsIter')
It is an Excel file where the user can apply the likelihood ratio test. The final log likelihood and the number of estimated parameters for both the restricted and the unrestricted models must be provided.
It is a Matlab code to compute the correlation structure of a cross-nested logit model.
It is an Excel sheet that allows to compute the variance of the difference and the ratio of two estimators. A typical application in the context of discrete choice is the computation of the standard error for parameters like value-of-time. The approximation computed here come from the Taylor series, where terms higher than second order are ignored. Source: MVA et al. (1998) "Value of Trave Time Savings".
biocheckdata [included in the distribution]
The syntax is: biocheckdata
mydata.dat
The script checks if a datafile
is complying with the requirements of biogeme. In
particular, it checks if the number of elements in
each row matches the number of headers in the first
row. It also detects if some entries are not
numeric.
/usr/local/bin/biocheckdata.
biopreparedata [included in the distribution]
The syntax is: biopreparedata mydata.csv
The script prepares a CSV data file in a format requested by biogeme. Each column containing strings is coded with numbers. The following conventions are adopted:
- Strings are delimited with double quotes.
"
- Each blank in the name of a header is replaced by a underscore.
_
- Entries are separated by a comma.
,
- If an entry of the first row is numeric, the corresponding column is supposed to contain only numerical values. If a non-numerical value is detected in another row, is it replaced by
99999.
/usr/local/share/biogeme/biopreparedata.py in order to change these conventions.
If an entry in the first column is a string, the script associates each string in the corresponding column with a numeric value.
Two files are generated:
biogeme_mydata.csvis the data file complying with the biogeme requirements.legend_mydata.csvdescribes the codings that have been used for non numeric data.
Suppose that the mydata.csv file contains
Id, The name, The rank 1, "Me", 2 2, "You", 3 3, "Him", 3Then, the generated file
biogeme_mydata.csv will contain
Id ___The_name _The_rank 1 2 2 2 0 3 3 1 3and the generated file
legend_mydata.csv will contain
+++++++++++++++++++++++++ Legend for column ___The_name +++++++++++++++++++++++++ 0 : "You" 1 : "Him" 2 : "Me"
Tool to organize a list of raw numbers into categories to plot an
histogram using bins of a given size. The raw data should contain only a list of numbers. The
syntax is
histogram filename binSize.
For example, if the Raw data file is
0.2748861875 1.2194215178 -0.1088626369 1.887765541 0.143842688 0.5121648584 0.9323810467 0.1969901739 -0.2501622963 0.6349579371 -0.6544964817 0.1684235135 1.0532380188 0.9794024028 0.8001565071 -0.7680558349 -2.5749162274 0.8619355768 -0.0267481139 -0.9200574846the command is
histogram test.dat 1.0, and the following files are
generated: _hist_test.dat
Value Frequency -3.0 1 -1.0 6 0.0 10 1.0 3meaning that there is 1 value between -3 and -1, 6 values between -1 and 0, 10 values between 0 and 1, and 3 values larger than 1. A Gnuplot file
_hist.gp is also generated
set style data histogram set style histogram cluster gap 0 set style fill solid 1.0 plot '_hist_test.dat' using 1:2 ti col smooth frequency with boxes
A weighted version of the tool exists. Instead of couting the
number of value in each bin, it adds the weights of these
values. The raw data file must contain two values on each row: the
value, and its weight. The
syntax is
weightedhistogram filename binSize.
For example, if the Raw data file is
0.2748861875 0.9 1.2194215178 0.2 -0.1088626369 1.0 1.887765541 1.0 0.143842688 0.6 0.5121648584 0.1 0.9323810467 0.1 0.1969901739 0.4 -0.2501622963 0.3 0.6349579371 0.2 -0.6544964817 0.4 0.1684235135 0.1 1.0532380188 0.8 0.9794024028 0.9 0.8001565071 0.3 -0.7680558349 0.8 -2.5749162274 0.4 0.8619355768 0.1 -0.0267481139 0.3 -0.9200574846 1.0the command is
weightedhistogram test.dat 1.0, and the following files are
generated: _hist_test.dat
Value Frequency -3.0 0.4 -1.0 3.8 0.0 3.7 1.0 2.0meaning that the total weight of values between -3 and -1 is 0.4, the total weight for values between -1 and 0 is 3.8, the total weighr for values between 0 and 1 is 3.7, and the total weight for values larger than 1 is 2. A Gnuplot file
_hist.gp is also generated
set style data histogram set style histogram cluster gap 0 set style fill solid 1.0 plot '_hist_test.dat' using 2:xtic(1) t '_hist_test.dat'
xml-file for code highlighting of Biogeme .mod-files in notepad, and MS Excel sheet for Horowitz-Test (non-nested hypothesis).