The Price of Big Science
uncertainties that are often associated with model parameters . Even though many parameters in system dynamics models represent quantities , it is very difficult or even impossible to measure the parameters to a high degree of accuracy in the real world . Also , some parameter values change in the real world over time ( Forrester 1996 ). In this regard , our system dynamics model was tested to see how published article growth behavior in our system dynamics model would change after altering key parameters . The sensitivity analysis also validates the model . For analyzing sensitivity in our system dynamics , two key parameters were analyzed for the sensitivity of published article growth based on average lifetime and creation rate . The sensitivity of each parameter change was analyzed by endowing each parameter with a range as follows :
average lifetime = RANDOM UNIFORM ( 7 , 8 ) creation rate = RANDOM UNIFORM ( 0.30 , 0.35 )
where the average lifetime is assigned between seven and eight years ( as found in the real world ), and the growth rate is randomly unit-distributed between the fraction per year 0.30 and 0.35 . Also , where : ( 1 ) Change current ( fractional ) creation rate , 0.325 up to 0.35 or down to 0.30 in the year 2009 to see what the published article growth would be in future . Growth patterns of published articles appeared as seen in Figure 8 . When the creation rate was up to 0.35 , the number of published articles became 5.774M ; on the other hand , when the creation rate was down to 0.30 , the number of published articles became 1.671M . Thus , this output shows that published article growth could be more sensitive when it was down than when it was up . We can see that the scientific knowledge output growth would be considerably daunted when creation rate was down , but it still shows a growth pattern .
( 2 ) Change the current parameter value regarding average lifetime , 7.5 years up to 8.0 years , or down to 7.0 years in the year 2009 to see what the published article growth would be in the future . Growth patterns of published articles appeared as seen in Figure 9 . When the average lifetime was up to eight years , the number of published articles became 6.737M ; on the other hand , when the average lifetime was down to seven years , the number of published articles became 1.905M . This output shows that published article growth could be more sensitive when it is up than when it is down .
( 3 ) Conduct sensitivity analysis by considering changes of both creation rate and average lifetime . The graph in Figure 9 shows the behavior of published article growth when the two parameters are changed at the same time .
The current extended simulation line is based on the parameter values creation rate = 0.35 , and average lifetime = 7.5 . At the extended simulation line , there exist four confidence bounds ( 50 %, 75 %, 95 %, and 100 %) for all the output values of published article growth shown in Figure 10 . These bounds were generated when the two parameters were randomly varied around their distributions at the same time . From this sensitivity analysis , we can expect that the growth of published articles would show a purely exponential growth rate , especially when each parameter was changed to an upper value . However , when the parameters , creation rate and average lifetime were lowered below the current parameter values , the published article growth does not follow a purely exponential growth pattern .
121