��������������������������
example , a model to help manage sexually transmitted diseases must include sexual contact patterns . As with the personal experience examples , it is not practical to construct cohorts for all the different combinations of states and estimate transition rates .
In these situations it is also much more natural to model the heterogeneity directly by using individuals instead of cohorts . Individual modeling allows the spatial or network structure to be explicitly included in the model : individuals are ‘ at ’ a location or pairs of individuals are connected in the network . Common techniques are Agent Based Modeling and , if the network is important , Social Network Analysis .
Individual modeling is also appropriate where certain states or combinations of states are rare , so as to avoid rounding problems . The number of people represented in a relatively rare cohort may be very small and any behavior that occurs for some fraction of the cohort could lead to ‘ part ’ individuals . For example , if the annual death rate is 0.2 for a specific age group , a cohort model can reduce the cohort size by 20 % during a year timestep , but an individual model will instead calculate a random number for each person and ‘ kill ’ that person if their random number is less than 0.2 . For large cohorts , the two approaches will lead to similar results . However , consider a cohort comprising only three people , sometimes there will be 0 , 1 , 2 , or even 3 deaths . In any case , there won ’ t be 0.6 deaths ; the result under the proportion approach , and rounding to 1 death each timestep would result in too many deaths during the simulation . Individually simulated State Transition techniques and Dynamic Microsimulation are effective for dealing with these situations .
�������������������������������
Complex policy issues typically involve considerable interaction between the entities within the system . After all , complex behavior arises from interactions rather than the behavior of the entities individually . To reproduce the complex behavior , the modeling technique must be able to take into account the form of those interactions as well as the entities involved . For static models , such taking into account may be simply documenting or describing the interactions . Dynamic models , however , must include the influence of interactions when simulating the system ’ s behavior through the passage of time .
There are three broad levels of interaction . Some behaviors will be independent ; that is , there is no interaction . Indirect interaction arises where the behaviors of one entity affect the behaviors of some other entity without any contact . Typically this occurs by imposing constraints and hence changing their options , for example , by making resources unavailable . Direct interaction is more explicit ; the behavior of one entity in the model is affected by the actions of some other entity with which it has contact . Interaction may occur between entities of the same type or different types .
Two of the simplest models to show complex behavior concern population growth . Consider a population of newly introduced rabbits in a large grassland where there is no natural predator . Each rabbit has some age dependent chance of dying each year and , if female , some other chance of giving birth . Conceptualized simply as fixed probabilities , each rabbit ’ s life path is independent of all the other animals ; that is , it is not affected by the number or activity of other animals . Eventually , however , the grassland will be overrun with rabbits and
81