MerQuR+Statistics Glossary

Statistics Glossary → Abbott correction (natural response)

Abbott correction (natural response)concept

A method used in dose-response experiments to correct the response rate due to treatment by accounting for responses that occur even without treatment (such as natural deaths in the control group).

In an insecticide trial, some insects in the control group also die from old age or disease. If these natural deaths are ignored, the insecticide's effect looks bigger than it is. The Abbott correction assumes the observed response rate has two sources: P_observed = C + (1 - C)·P_treatment, where C is the natural response rate. The corrected response is then (P_observed - C) / (1 - C), the proportion killed by the treatment among those that would not have died naturally.

In probit or logit dose-response models, C can be entered as a fixed value or estimated by maximum likelihood together with the other parameters; the latter carries the uncertainty in the control group into the model. Effective doses (such as the ED50) are computed from this corrected model. If the natural response rate is high (roughly above 20%), the correction becomes unreliable and the trial conditions should be questioned. The method assumes the natural response is independent of the treatment.

An everyday example

In a laboratory, mosquitoes are given different doses of an insecticide. In the untreated control group, 10% of the mosquitoes die within 24 hours anyway. If 55% mortality is seen at the highest dose, the Abbott-corrected mortality due to the insecticide is (0.55 - 0.10) / (1 - 0.10) = 0.50, that is 50%.

In MerQur

In Probit Regression, the 'Dose-response options: Natural response' option fits the model P = C + (1 - C)·Φ(xb); C is estimated by maximum likelihood and effective doses are computed from this model.

Analyses where this concept appears

Probit Regression

Related concepts

Effective dose (ED50, LD50) and dose-response