Poisson Rate Calculator

Estimate rate from observed events.

Please enter a valid number of events (≥ 0)

Examples

Enter event count and time period to estimate rate

What Is Poisson Rate Estimation?

Given observed events, we estimate λ (lambda)—the true average rate. This calculator provides the rate estimate and its confidence interval.

Point estimate: λ̂ = (total events) / (total exposure)

Example: If 15 events occurred over 5 time periods: λ̂ = 15/5 = 3 events per period

But how precise is this estimate? The confidence interval answers this.

Confidence Interval for λ

For observed count X over exposure t:

Exact method (Garwood):

Lower: χ²(α/2, 2X) / (2t) Upper: χ²(1-α/2, 2(X+1)) / (2t)

Where χ² is the chi-square inverse function.

Normal approximation (large λ̂):

CI = λ̂ ± z* × √(λ̂/t)

For 95% CI: z* = 1.96

Examples

Medical Research

"In 1000 person-years of follow-up, 25 events occurred. What's the rate and 95% CI?"

λ̂ = 25/1000 = 0.025 per person-year = 25 per 1000 person-years 95% CI: approximately 16 to 37 per 1000 person-years

Quality Control

"Over 100 production runs, 340 defects were found. Estimate the defect rate."

λ̂ = 340/100 = 3.4 defects per run 95% CI: approximately 3.0 to 3.8 defects per run

Traffic Analysis

"8 accidents occurred at an intersection over 3 years. Estimate the annual rate."

λ̂ = 8/3 = 2.67 accidents per year 95% CI: approximately 1.2 to 5.3 accidents per year

Website Analytics

"A server logged 450 errors in 30 days. Estimate errors per day."

λ̂ = 450/30 = 15 errors per day 95% CI: approximately 13.6 to 16.5 errors per day

Point Estimate Properties

The sample rate λ̂ = X/t is:

Unbiased: E(λ̂) = λ Consistent: λ̂ → λ as exposure increases Maximum likelihood: λ̂ is the MLE for λ Efficient: λ̂ achieves minimum variance among unbiased estimators

Variance: Var(λ̂) = λ/t

Standard error: SE(λ̂) = √(λ̂/t)

Different Exposure Scenarios

Single Observation Period

X events in t time units: λ̂ = X/t

Multiple Observation Periods

X₁ events in t₁, X₂ events in t₂, etc.: λ̂ = (X₁ + X₂ + ...) / (t₁ + t₂ + ...)

Person-Time Exposure

Common in epidemiology: λ̂ = (total events) / (total person-years at risk)

Applications

Epidemiology

Incidence rates measure disease occurrence:

  • Incidence rate = (new cases) / (person-time at risk)
  • 95% CI allows comparison across populations

Reliability Engineering

Failure rates quantify component reliability:

  • Failure rate = (failures) / (total operating time)
  • Mean time between failures = 1/λ

Process Improvement

Track defect rates over time:

  • Baseline rate vs. post-intervention rate
  • Confidence intervals reveal whether improvement is significant

Insurance

Claim rates inform pricing:

  • Claims per policy-year
  • Confidence intervals account for sampling uncertainty

Confidence Interval Interpretation

"λ̂ = 3.2 defects per hour, 95% CI: [2.1, 4.6]"

We're 95% confident the true rate falls between 2.1 and 4.6 defects per hour.

If comparing to a target (e.g., 5 defects/hour):

  • If 5 is above the CI upper bound → rate is significantly below target
  • If 5 is within the CI → can't distinguish from target

Comparing Two Rates

For two groups with rates λ₁ and λ₂:

Rate ratio: RR = λ₁/λ₂

Confidence interval for log(RR): SE(log(RR)) = √(1/X₁ + 1/X₂)

If the CI for RR excludes 1, the rates differ significantly.

When Zero Events Are Observed

If X = 0:

  • Point estimate: λ̂ = 0
  • 95% CI upper bound: 3/t (approximately)
  • This is the "Rule of Three"

Example: No failures in 100 hours → 95% CI upper bound ≈ 3/100 = 0.03 per hour

Assumptions and Limitations

Constant Rate

The Poisson model assumes λ doesn't change over time. If rates vary:

  • Split analysis by time period
  • Use more complex models (non-homogeneous Poisson)

Independent Events

Events must be independent. Clustering (events triggering other events) violates this.

Known Exposure

Exposure time must be accurately measured. Uncertainty in exposure adds to overall uncertainty.

Historical Context

Frank Garwood (1936) developed the exact confidence interval for Poisson rates using the chi-square distribution relationship.

The method became standard in epidemiology and reliability engineering, where rate estimation is fundamental.

Modern software computes exact intervals routinely, but the chi-square relationship remains the foundation.

Related Calculations

Summary

Poisson rate estimation converts observed event counts into rate estimates with uncertainty quantified by confidence intervals. The point estimate λ̂ = events/exposure is intuitive, while the confidence interval (using the chi-square relationship or normal approximation) reveals precision. From epidemiological incidence rates to engineering failure rates to quality control metrics, Poisson rate estimation is fundamental to understanding event frequencies and making comparisons across groups or time periods.

Related Tools