residual analysis-based chi-square detection method  (CH Instruments)

Journal: PLOS Pathogens

Article Title: Prior exposure to pathogens augments host heterogeneity in susceptibility and has key epidemiological consequences

doi: 10.1371/journal.ppat.1012092

Figure Lengend Snippet: Host susceptibility distributions for house finches from variable prior exposure treatments: no prior exposure (A,B); low-dose (C); or high-dose (D). Colored lines show estimated susceptibility distributions from either homogeneously (A) or gamma-distributed (B-D) models (note distinct axes for the two models). In (A), host infection probability per 1000 bacterial particles ( p ) is shown as the single best fit parameter p (dotted vertical lines represent 1 standard error) for the homogeneous model, which was the best fit model for the no prior exposure group (see ). In (B-D), the best fit parameters (shape and scale) for gamma distributions (teal lines) are listed for each group, and vertical gray lines indicate mean susceptibility ( x ) for that treatment. Lighter shading represents 95% confidence regions for gamma distributions, obtained by bootstrapping chi-squared residuals to create 1,000 pseudoreplicates of infection data and then refitting the model to pseudoreplicates, as per [ , ]. The gamma model was the best fit for only the low-dose and high-dose groups. Gamma estimates are also shown for the no prior exposure group (B) because this allowed more equivalent comparisons for certain SIR simulations (see ).

Article Snippet: We simulated the heterogeneous and homogeneous models using the parameter estimates obtained from bootstrapping chi-squared residuals (see ).

Techniques: Infection