Small confidence intervals in underdispersed data (Delta Method) about biodosetools HOT 3 CLOSED

@dendesfelder Would this also apply for translocations and micronuclei? Or should we only use the empirical error for them?

from biodosetools.

I'm trying this implementation at the moment:

disp <- case_data[["DI"]]

if (disp >= 1) {
  # Use empirical error sqrt(var / N) if disp >= 1
  if (aberr_module == "dicentrics" | aberr_module == "micronuclei") {
    lambda_est_sd <- case_data[["y_err"]]
  } else if (aberr_module == "translocations") {
    lambda_est_sd <- case_data[["Fg_err"]]
  }
} else {
  # Use Poisson error if disp < 1
  lambda_est_sd <- sqrt(case_data[["X"]]) / case_data[["N"]]
}

Would these results make sense? Using C0=5, and C1=2 and the fitting curve from IAEA:

Before fix	After fix

from biodosetools.

Seems to make sense. Which calibration curve coefficients did you use for dose estimation. I would say that the same applies for translocations and Micronuclei, too. For very high doses of low-LET we would expect some degree of underdispersion and we might therefore slightly overestimate the uncertainty with this approach. However, I believe it's better to overestimate than underestimate.

from biodosetools.