Skip to content

Statistical methods

All statistical methods are documented in the source code, which is publicly available. https://github.com/shapiromatron/hawc. If you're interested in locating a particular method and have troubles navigating the source code, please contact us and we can help.

Bioassay

  • For single dose-response confidence intervals displayed on the dose-response plot:
  • For continuous data, 95% confidence intervals are calculated using a two-tailed t-test (eg., https://hawcproject.org/ani/endpoint/1/)

  • For dichotomous data, intervals are calculated using the inverse standard normal cumulative distribution function evaluated at \(1-\alpha/2\), using a two-sided 95% confidence interval (eg., https://hawcproject.org/ani/endpoint/30/)

  • For calculating percent control relative to the control response (in data-pivot, eg., https://hawcproject.org/summary/data-pivot/assessment/1/feature-example-percent-difference/):

    • 95% confidence intervals are calculated using a Fisher Information Matrix, assuming independent normal distributions
    • This statistic assumes that first dose-group is control, and all other dose-groups are compared to control
    • Calculation of confidence intervals requires the following:
      • The endpoint dataset must provide variance measures (i.e. standard deviation or standard error)
      • The control mean value must not be equal to zero (cannot divide by zero)
      • The n for each dose-group must be provided, including control
    • A detailed description is available: CI description

In-vitro

  • For single dose-response confidence intervals displayed on the dose-response plot:

  • same as continuous methods for bioassay data, as described above

  • For calculating percent control relative to the control response (in data-pivot):

    • same as methods for bioassay data, as described above

Epidemiology (v1)

  • For confidence intervals on epidemiology result-datasets

  • For calculating percent control relative to the control response (in data-pivot):

    • same as methods for bioassay data, as described above
    • The control group is determined by the following rules:
      • If 0 groups are marked control (using comparison-set groups), the first group will be chosen as control
      • If 1 groups is marked control (using comparison-set groups), the first group will be chosen as control
      • If ≥ 2 groups are marked control (using comparison-set groups), a random control group will be chosen for control for all calculations