User:Elisabetta Lambertini/Notebook/Salmonella/2010/11/01

Project name

Main project page Previous entry      Next entry

Risk from Salmonella in almonds Debugging. Slight change of plans on how to look at outcomes. Danyluk 2006 used the geometric mean of #cases/yr. It makes sense since the distribution of both Prob(ill) and #cases/yr is highly skewed. Note that zeros are seen as errors when taking the Log and eliminated from the vector, so the distribution of LogCases/yr has less data points than #cases/yr. This doesn't change the result. Outbreak scenario: we expect a result of ~ 1500-2000 cases (93 reported cases * (100/20), assuming a reporting rate of 20%. Test: point estimate across the range of back-calculated concentrations: from 200 MPN/100 g, to 444 MPN/100 g; "high" distribution, using detection limit as lower boundary (1 CFU/100 ml to have + enrichment, back calculated as ~ 56 MPN/100 g) "low" distribution using 0.1 as lower boundary. Note: MPN distribution fitted as cumulative. Assumptions: the whole amount shipped btw feb and april 2001 was contaminated at the same level as the 26 samples analyzed (also try with just the amount of lots C+D)