Skip to main content
. Author manuscript; available in PMC: 2010 Jul 30.
Published in final edited form as: Trop Med Int Health. 2006 Feb;11(2):188–196. doi: 10.1111/j.1365-3156.2005.01555.x

Table 2.

Predictive accuracy metrics for the adjusted and unadjusted models of access and utilisation of GoK–MoH health services for the treatment of paediatric fevers

Model Overall Accuracy† Kappa κ‡ P-value 95% Confidence Interval
Euclidean 0.72 0.71 0.00 0.70–0.72
Transport network-not adjusted for competition 0.74 0.73 0.00 0.73–0.74
Transport network-adjusted for competition 0.84 0.83 0.00 0.79–0.87
†

Overall accuracy: this is simply the number of correctly classified sample points divided by the total number of sample points

‡

Kappa statistic (κ): is the chance-corrected measure of model accuracy, based on the actual agreement between predicted and observed values and the chance agreement between the row and column totals for each classification (Congalton & Green 1999). P-values reflect the probability that a model performs better than random chance at predicting the choice of a health facility