A. List of Model Parameters. B. Model Technical Description

Appendix-A: List of Model Parameters
291

In this appendix, we present a list of model parameters and their default values
including assumed constants that require validation as future work. These
parameters are in addition to the model description and behavioural rules that
we have discussed in Chapter 4. Table 41 presents values of the model
parameters that have are used as default in the simulation experiments
reported in Chapter 5 of this thesis. The far right column indicates the source of
these values: ‘LT’ refers to values adapted from external references and
sources and ‘DE’ means that our case study partners and other domain experts
have suggested the respective values. Finally, ‘UV’ corresponds to unvalidated
assumptions that have been introduced in the model in the absence of available
evidence during the course of this research. We hope that these values may be
validated or replaced with better-informed guesses following further evidencegathering
processes.
292

Table 41: List of model parameters and their default values/ranges: ‘LT’: values adapted from other sources; ‘DE’: input from domain
experts/case study partners; ‘UV’: unvalidated assumptions due to lack of evidence
Parameter Default value Description Status
General Setting
churchParticipationRate 0.8 Proportion of households being member of a church
clusterProportion 25 Number of households per moiety
healthExpenditure True Flag to turn health expenditure: on (true) off (false)
HIV True Flag to turn HIV spread in the simulation: on (true) off (false)
initialFuneralClubs 9 Number of initial funeral clubs
initMigratedAgents 0.03 Initial proportion of migrant agents
networkRecordInterval 10 Interval at which the generated networks are recorded
numDenominations 4 Number of church denomination
numHouseholds 150 Initial number of households
seed 1199995797250l Seed for the random number generators
showGUI False Flag to turn the GUI on (true) or off (false)
Agent Related
agentLoseJobChance 0.01 Chance for an migrant agent to lose job UV
agentStopMigrationAge 55 Age at which an individual agent stop LT
amountReqdMigration 1000 Rand Amount required for migration for an agent in Rand DE
birthrate 23 Birth rate per 1000 person per year LT
chanceMigrationFemale 0.03 Chance for migration for male agent DE
chanceMigrationMale 0.15 Chance for migration for male agent DE
femaleActiveSexAge [13, 45] years Age range for sexually active female agents that participate in courtship LT
fertilityAgeLimit Normal (40, 5) Fertility age limit for a female agent sampled from Normal distribution LT
friendshipLimit 7 Upper limit for number of friends for an individual agent UV
infantMortalityRate 52 Infant mortality rate per 1000 person per year LT
lobola 10,000 Rand Maximum amount for lobola DE
maleActiveSexAge [16, 55] years Age range for sexually active male agents that participate in courtship LT
maxCourtshipDuration Lognormal(0.9, 0.7) Maximum dating time for an agent sampled from truncated lognormal distribution within range [12, 18] months LT
maxSexPartnerFemale Lognormal(0.2, 0.3) Maximum sex partners for a female agent at a time sample sampled from Lognormal distribution LT
maxSexPartnerMale Lognormal(0.4, 0.7) Maximum sex partners for a male agent at a time sample sampled from Lognormal distribution LT
migrantVisitFreq Uniform(3, 5) Frequency of visit of migrant agents to home LT
numberofMigrationTries Uniform(3, 9) Number of months that an agent waits until it attempts to migrate again UV
randomParnerProb 0.05 Probability for picking a random partner UV
singelRemittanceLimit Uniform(24, 48) Number of months for an unmarried agent to send remittances; range sampled from Uniform distribution UV
chanceLooseJob 0.01 Chance for a migrant agent to lose job UV
293

Parameter Default value Description Status
Mining Related
miningStartTime 61 Mining starts at the 5 th year UV
outsiderAgeRange Normal(29, 5) Age range for outsider agents; sampled from truncated Normal with range [16, 40] years LT
outsiderMineTrainProb 0.001 Chance that an outsider agent is an skilled mineworker UV
outsiderWaitingTime Uniform(1, 5) Number of months that an outsider agent waits without a job at mine UV
skilledCapacity 100 Maximum number of skilled agents that can be employed at the mine UV
skilledWorkerSalary Normal(5000, 500) Monthly salary for skilled workers at the mine sampled from Normal distribution DE
unskilledCapacity 350 Maximum number of unskilled agents that can be employed at the mine UV
unskilledWorkerSalary Normal(1000, 250) Monthly salary for unskilled workers at the mine sampled from Normal distribution DE
Funeral Clubs Related
fc_GracePeriod 6 months Grace period for members without paying dues UV
fcCapacity Uniform(11, 19) Maximum number of funeral club members per club in the model UV
fcInitialMoney 10,000 Rand Initial money assigned to funeral clubs UV
funeralClubDue 70 Rand Monthly funeral club fee paid by members DE
maxFC_Cost 7500 Rand Maximum funeral cost for a bereaved household DE
midFC_Cost 5000 Rand Average funeral cost for a bereaved household DE
minFC_Cost 2500 Rand Minimum funeral cost for a bereaved household DE
Savings Clubs (‘stokvel’) Related
minStokvelMembers 3 Minimum number of members required to form a savings club UV
probFacilitator 0.20 Chance for an agent to act as a facilitator for initiating a savings club UV
stokvelFee 50 Rand Monthly savings club fee paid by members DE
Household Related
chanceBorrowFood 0.20 Chance for borrowing food in a month time UV
chancePieceJob 0.75 Chance for an eligible household to announce piecewise jobs UV
educationFee 20 Rand Monthly school fee per child agent UV
healthExp 50 Rand Health expenditure per member of a household UV
maxMoneyToFeed 165 Rand Maximum money required to feed an adult male DE
maxStartWealth 2500 Rand Maximum starting wealth for households DE
minMoneyToFeed 120 Rand Minimum money required to feed an adult male DE
minStartWealth 1500 Rand Minimum starting wealth for households DE
numPiecewiseJobs Uniform(1, 3) Number of piecewise jobs announced by an eligible households UV
richHHCriteria 15,000 Rand Minimum wealth for a household required to announced piecewise jobs at household UV
294

Appendix-B: Model Technical Description
295

The model is developed in Java and uses the Repast 3.1 Simulation toolkit
(North et al. 2001). Figure 117 shows the UML diagram for the classes
implemented in the model and those inherited from the Repast libraries. The
source code along with the documentation is available in the CD attached to
this thesis. It is also publicly available for download from the CAVES Project
Portal – http://cfpm.org/caves/CAVESWiki/Models
(last accessed: 25/06/2008).
Figure 117: UML class diagram showing the class relations and hierarchy in the model
296

Running the model from the Repast
Since the model is using Repast for its simulation infrastructure, simulation runs
can be started from the Repast graphical user interface (GUI). Invoking the
main() method of the model class (display.SouthAfricanModel) will bring up the
Repast GUI (see the screenshot in Figure 118). One can change model settings
from the GUI and start, pause or stop simulation runs from there.
Figure 118: Screenshot of the model with the GUI
Running the model from console as a batch run
To save on time and memory it is possible to run the model from console. This
is a so-called batch run in Repast terminology and is started by invoking the
Repast class SimInit with the model class as parameter:
java uchicago.src.sim.engine.SimInit -b display.SouthAfricanModel
Additional Java libraries
The model requires the following libraries:
• flanagan.jar - Michael Thomas Flanagan's Java Scientific Library
(last accessed: 25/06/2008)
297

• repast.jar – Repast Simulation Toolkit 3.1
(last accessed:
25/06/2008)
• colt.jar – CERN Library for High Performance Scientific and Technical
Computing in Java
(last accessed: 25/06/2008)
Setting the classpath of the above libraries is necessary to run the model main
class.
Model Output
The model produces output on several levels:
GUI: friendship network and stokvel network (updated dynamically), extended
family network, social links of the households in the village
Files: complete model output. For each model run a folder is created in the
specified output
At every interval: networkRecordInterval, the sexual network, friendship network
and households social links are output in Pajek format 109 .
Time series data for labour activity, demographics, HIV/AIDS incidence and
prevalence are output at every time step.
109 Courtesy: Friedrich Krebs, CESR, Kassel
298