Kathryn Alexander
Thursday, October 31, 2013 - 12:16pm

The current version of the CRISP package includes data up until August 2013. Forecasts here are for October 2013.

These hierarchical model forecasts are for the five EOIs in the current CRISP package version. Listed countries are either the three with the highest risk for the respective EOI in October, or all of the countries with a probability of occurrence higher than 0.9. Only countries that did not have an actual incidence of the EOI of interest are listed in the respective watch list.



Florian Hollenbach
Thursday, October 31, 2013 - 12:16pm

The current version of the CRISP package includes data up until May 2013. The models are based on an in-sample calibration period up until September 2012 and a test period from October 2012 through May 2013. Based on the statistical models we then make forecasts for 6 months into the future, here we focus on our forecasts for the five EOIs for July 2013.



Mike Ward
Thursday, October 31, 2013 - 12:16pm

This isn't about the conflict we normally study, but Florian Hollenbach, Jacob Montegomery and I have developed some techniques for averaging forecasts. The basic idea was published in Political Analysis (Summer 2012), but we have been involved in predicting the popular vote for the incumbent in the upcoming election. Our approach takes the models and estimates from experts in American Politics. We then create a weighted ensemble of those estimates, where the weights are established based on how accurate the models have been in past predictions.



Mike Ward
Thursday, October 31, 2013 - 12:16pm

We have some new input data which  in combination with our current models, provide some interesting predictions of new crisis events.  In this example, I'll focus on the variable international crisis, but provide a table with a broader set of predictions as well.



Mike Ward
Thursday, September 27, 2012 - 11:37am

Based on the analysis and data as of 30 August 2012 (build 0.61) the following results are of interest.

We have recently incorporated data into our models that brings us current to the middle of 2012. And as a result, we have calculated a set of watch lists for countries that show a high probability of the onset of a new event in the next 6 months, based on our analysis of data from January 1997 through June 2012.