In this work we studied the relationship between dengue incidence in Cali and the climatic variables that are known to have an impact on the mosquito and were available (precipitation, relative humidity, minimum, mean, and maximum temperature). Since the natural processes of the mosquito imply that any changes on climatic variables need some time to be visible on the dengue incidence, a lagged correlation analysis was done in order to choose the predictor variables of count regression models. A Principal Component Analysis was done to reduce dimensionality and study the correlation among the climatic variables. Finally, aiming to predict the monthly dengue incidence, three different regression models were constructed and compared using de Akaike information criterion. The best model was the negative binomial regression model, and the predictor variables were mean temperature with a 3-month lag and mean temperature with a 5-month lag as well as their interaction. The other variables were not significant on the models. And interesting conclusion was that according to the coefficients of the regression model, a 1°C increase in the monthly mean temperature will reflect as a 45% increase in dengue incidence after 3 months. The rises to a 64% increase after 5 months.

Monthly dengue incidence data was provided by the Public Health Department of Cali. The climatic data was collected from the Hydrology, Meteorology and Environmental Studies Institute (IDEAM) webpage. Missing data was imputed using a random forest regression model.