For several countries with disparities between Cascade Indicators it is tricky to identify a good fit between the model and the data.
For example, in Nigeria, we have a very high PLHIV value between 2010 and 2015, but we think that PLHIV Diagnosed is very low. Indeed, we have data on this in 2012, but nothing to seed the model with in 2010. Therefore, due to the lack of data, the model estimates that the number of PLHIV Diagnosed in 2010 is somewhere between the number on ART in that year (v. low) and the number of PLHIV (v. high) - sometimes it hits the right mark, but often produces results with VERY high PLHIV diagnosed values.
We need an 'alternate' calibration switch in the model that uses optim() to identify one set of parameters that fit all of the data and is a better fit to data.
jackolney
commented
7 years ago
For several countries with disparities between Cascade Indicators it is tricky to identify a good fit between the model and the data.
For example, in Nigeria, we have a very high PLHIV value between 2010 and 2015, but we think that PLHIV Diagnosed is very low. Indeed, we have data on this in 2012, but nothing to seed the model with in 2010. Therefore, due to the lack of data, the model estimates that the number of PLHIV Diagnosed in 2010 is somewhere between the number on ART in that year (v. low) and the number of PLHIV (v. high) - sometimes it hits the right mark, but often produces results with VERY high PLHIV diagnosed values.
We need an 'alternate' calibration switch in the model that uses
optim()to identify one set of parameters that fit all of the data and is a better fit to data.This is to be attempted post-croatia.