1
PROBIT ANALYSIS
Definition
Probit analysis is a statistical method used to model the relationship between a binary outcome (like
success or failure, presence or absence) and one or more independent variables. It transforms a sigmoid
dose-response curve into a straight line, making it easier to analyze the relationship using regression
techniques. Essentially, it's a type of regression analysis where the dependent variable has only two
possible values.
Quick overview
• Probit analysis is a type of regression used to analyze binomial response variables.
• It transforms the sigmoid dose-response curve to a straight line that can then be analyzed by
regression either through least squares or maximum likelihood.
• Probit analysis can be conducted by one of three techniques:
o Using tables to estimate the probits and fitting the relationship by eye,
o Hand calculating the probits, regression coefficient, and confidence intervals, or
o Having a statistical package such as SPSS do it all for you.
Key characteristics of probit analysis
Here are the key characteristics of probit analysis, especially as applied in biological and toxicological
research:
1. Binary response variable
Probit analysis is designed for binary outcomes (e.g., alive vs. dead, success vs. failure, diseased
vs. healthy).
The data reflect the proportion of individuals exhibiting a particular response at each dose level.
2. Dose-response relationship
It models how increasing doses or concentrations of a substance influence the probability of a
particular outcome.
3. Nonlinear relationship linearized by transformation
The raw dose-response curve is usually sigmoidal (S-shaped). Probit analysis transforms the
response (percentage) using the inverse standard normal distribution (z-score) to make the
relationship linear for regression.
4. Uses normal distribution assumptions
Probit analysis assumes the underlying distribution of tolerance (or susceptibility) in the
population is normal.
The cumulative response is modeled using the cumulative normal distribution function.
5. Enables estimation of LD50, LC50, ED50
One of the main uses is to determine-
o LD50 (Lethal Dose for 50% of population)
o LC50 (Lethal Concentration for 50%)
o ED50 (Effective Dose for 50%)
These values are interpreted from where the probit = 5.
6. Straight-line model
After transformation, a linear regression is applied:
2
𝑃𝑟𝑜𝑏𝑖𝑡 𝑦 = 𝑎 + 𝑏. (𝐷𝑜𝑠𝑒)
Sometimes dose is log-transformed for better model fit.
7. Handles multiple dose levels
It can accommodate multiple groups with varying doses to estimate a full dose-response
relationship.
8. Requires a range of responses
To be effective, data should include responses ranging from low to high mortality/effect (e.g.,
10% to 90%).
Avoids flat or step-like data which can distort model fitting.
9. Provides confidence intervals
Probit analysis provides 95% confidence intervals around LD50 or LC50 estimates, useful for
assessing the precision and reliability of results.
Key applications in biological science
Probit analysis has a broad range of applications in biological sciences, particularly in the fields of
toxicology, pharmacology, ecology, and genetics. Here are some key applications of probit analysis in
biological science:
Toxicology: Estimating lethal doses (LD50) or lethal concentrations (LC50) of toxic substances.
Pharmacology: Analyzing dose-response curves for drugs and pharmaceuticals.
Ecology: Assessing environmental pollutants and their effects on species.
Genetics: Studying the impact of genetic mutations, gene knockdowns, and mutagens.
Insecticide resistance: Evaluating resistance levels and effectiveness of insecticides.
Plant biology: Measuring the impact of herbicides and environmental stressors on plant survival
and growth.
Clinical research: Determining optimal drug dosages and vaccine efficacy.
Behavioral research: Quantifying the effect of stimuli on behavior and cognition.
How it works
Data collection: Experiments are designed to measure the probability of a response (like death) at
different doses of a stimulus.
Transformation: The probabilities are transformed into probit values using the inverse of the
cumulative standard normal distribution.
Regression analysis: A regression model is fitted to the transformed data, allowing for the analysis
of the relationship between the dose and the response.