Figure 3

Estimation of the proportion of individuals exposed to ARs (A), the proportion of mutants in rat population (B), the proportion of mutants in mouse population (C). A Generalized Linear Model as a smoothing model was used to analyze the geographical pattern. Binomial distribution for the response variable (proportion of rodent with ARs or proportion of mutated rodents) and logit link were used. The coordinates X and Y was used as explanatory variables. Proportion of rodent exposed to ARs increases from yellow to red. Proportion of mutant rodents increases from grey to black. White areas represent natural area where AVK treatments are unexpected. “0” symbols represent locations of sampled sites in natural area (control sites), “ + ” symbols, those in other locations. Note that map of mouse model (C) is adjusted on a smallest number of sites than (A) and (B) and that estimated coefficients (Xlong, Ylat and interaction between Xlong and Ylat) are less accurate.