STATISTICSRegressionStatistics Calculator
📊

Exponential Regression Calculator

Free exponential regression calculator. Fit y = ae^(bx) to data. R², doubling time, half-life, growt

Run CalculatorExplore data analysis and statistical calculations

Why This Statistical Analysis Matters

Why: Statistical calculator for analysis.

How: Enter inputs and compute results.

e
STATISTICSRegression

Exponential Regression — Log-Linear Fit

Fit y = ae^(bx) to data. R², doubling time, half-life, growth rate, prediction, residuals. Scatter + log-scale + residual charts.

Linear Regression →Cubic Regression →

Real-World Scenarios — Click to Load

Model Configuration

xy
exp_regression_results.sh
CALCULATED
$ exp_reg --model="e-based" --pairs=6
Coefficient a
100.0259
Coefficient b
0.4054
R²
1.0000
Model
ae^(bx)
Doubling time
1.7098
Growth rate
49.99%
Share:
Exponential Regression Result
y = 100.03 × e^(0.405x)
R² = 1.0000
Doubling: 1.71Growth: 50.0%
numbervibe.com/calculators/statistics/exponential-regression-calculator

Calculation Breakdown

LOG-LINEAR REGRESSION
Transform Y = ln(y)
Y = [4.605, 5.011, 5.416, 5.823, 6.227...]
FIT
Linear regression on (x, Y)
Y = 4.6054 + 0.4054x
ext{Least} ext{squares}: A = ext{intercept}, b = ext{slope}
Parameter a
100.0259
a = e^4.6054
Parameter b
0.4054
b = slope (continuous rate)
GOODNESS OF FIT
R² (original scale)
1.0000
R^{2} = 1 - SS_res/SS_tot
INTERPRETATION
Doubling time
1.7098
t_{2} = \text{ln}(2)/b
Growth rate r
49.99%
r = e^b - 1

Scatter with Exponential Fit

Log-Scale Plot (linear when fit is good)

Log-Linear Residuals

For educational and informational purposes only. Verify with a qualified professional.

Key Takeaways

  • Model: y = ae^(bx). Linearize: ln(y) = ln(a) + bx. Fit linear regression on (x, ln(y)).
  • Parameters: a = e^(intercept), b = slope. For y = ab^x, b_base = e^b.
  • R² computed on original scale: R² = 1 − Σ(yᵢ − ŷᵢ)² / Σ(yᵢ − ȳ)².
  • Doubling time: t₂ = ln(2)/b (growth, b > 0). Half-life: t₁/₂ = ln(2)/|b| (decay, b < 0).
  • Growth rate: r = e^b − 1 (as decimal). Multiply by 100 for percentage.
  • Prediction: ŷ = ae^(bx̂). Residuals: eᵢ = yᵢ − ŷᵢ.

Did You Know?

🦠Bacterial growth follows exponential curves in the log phase. Doubling time depends on temperature and nutrients.Source: Microbiology textbooks
☢️Radioactive decay is exponential. Half-life is constant regardless of initial amount — a fundamental property.Source: Nuclear physics
💰Compound interest is y = P(1+r)^t. Taking ln gives linear form: ln(y) = ln(P) + t×ln(1+r).Source: Finance
📈Population growth is often modeled as exponential in early stages. Carrying capacity leads to logistic curves.Source: Ecology
📊Inflation erodes purchasing power exponentially. A 3% annual rate doubles prices in ~23 years.Source: Economics
📱Viral adoption curves (e.g., social media) often show exponential growth before saturation.Source: Diffusion of innovations

Formulas

y = ae^(bx) ↔ ln(y) = ln(a) + bx

Linearize and fit Y = A + bx, then a = e^A, b = slope

Doubling time: t₂ = ln(2) / b = 0.693 / b

For growth (b > 0)

Half-life: t₁/₂ = ln(2) / |b|

For decay (b < 0)

R² = 1 − SS_res / SS_tot (on original y scale)

SS_res = Σ(yᵢ − ae^(bxᵢ))²

Log-Linear Regression Steps

1. Transform: Y = ln(y). Ensure all y > 0.

2. Fit linear regression: Y = A + bx using least squares.

3. Back-transform: a = e^A, b = slope. Model: ŷ = ae^(bx).

4. Compute R² on original scale (not on ln(y)) for interpretability.

5. Check residuals. Exponential fit assumes multiplicative errors; residuals may show patterns if wrong model.

Frequently Asked Questions

When should I use exponential vs linear regression?

Use exponential when the relationship shows constant relative growth (e.g., doubling every period). Linear is for constant absolute change.

What if some y values are zero or negative?

Exponential regression requires y > 0. Add a constant or use a different model (e.g., log-link GLM) for non-positive data.

How do I interpret the growth rate r = e^b − 1?

r is the per-unit increase in y as a proportion. r = 0.1 means 10% growth per unit x. For small b, r ≈ b.

Why is the log-scale plot linear?

ln(y) = ln(a) + bx is linear in x. So plotting (x, ln(y)) or (x, y) on log scale shows a straight line when the fit is good.

What is doubling time used for?

Doubling time = ln(2)/b. It is the time for y to double (growth) or halve (decay). Common in epidemiology, finance, and biology.

Model Comparison: y = ae^(bx) vs y = ab^x

y = ae^(bx): Natural exponential. b is the continuous growth rate. Doubling time = ln(2)/b.

y = ab^x: Discrete base. b is the growth factor per unit. Equivalent to ae^(x×ln(b)). Use when x is discrete (e.g., years, generations).

Limitations

  • • Assumes multiplicative errors. If errors are additive, consider nonlinear least squares.
  • • Log transform can amplify small errors. Outliers in y have large impact on ln(y).
  • • Extrapolation is risky. Exponential growth/decay cannot continue indefinitely in many real systems.
  • • R² on original scale may be lower than R² on log scale. Always report which scale you use.

Applications

Epidemiology

Early epidemic growth, doubling time, R₀ estimation.

Finance

Compound interest, investment growth, decay of purchasing power.

Chemistry

Radioactive decay, first-order kinetics, concentration vs time.

Physics

Newton's cooling, capacitor discharge, absorption.

Chart Interpretation

Scatter with fit: Data points and the fitted exponential curve. A good fit shows points closely following the curve.

Log-scale plot: On a log y-axis, exponential data appears linear. If the fit is good, points and line align along a straight line.

Residuals: Should be randomly scattered around zero. Patterns (e.g., U-shape) suggest a different model (e.g., logistic).

For growth data (b > 0), the scatter plot curves upward. For decay (b < 0), it curves downward. The log-scale plot linearizes both.

Worked Example: Bacterial Growth

Suppose you measure bacterial count at hours 0, 1, 2, 3, 4: y = [100, 165, 272, 449, 741]. Taking ln(y) gives [4.61, 5.11, 5.61, 6.11, 6.61] — roughly linear in x. Linear regression on (x, ln(y)) yields intercept ≈ 4.61, slope ≈ 0.5. So a = e^4.61 ≈ 100, b ≈ 0.5. Model: y = 100e^(0.5x). Doubling time = ln(2)/0.5 ≈ 1.39 hours.

For radioactive decay with half-life 5 years: b = −ln(2)/5 ≈ −0.139. The model y = N₀e^(−0.139t) describes the remaining amount.

When to Use Exponential vs Other Models

ModelBest For
ExponentialConstant relative growth/decay, doubling/halving
LinearConstant absolute change
LogisticS-shaped growth with saturation
Powery = ax^b, e.g., allometric scaling

Choose exponential when your data shows constant percentage change (e.g., 10% growth per year). Use logistic when growth saturates (e.g., market penetration).

Disclaimer: Exponential regression assumes y > 0. For data with zeros, consider adding a small constant or using a generalized linear model. Always check residuals.

👈 START HERE
⬅️Jump in and explore the concept!
AI

Related Calculators

Cubic Regression Calculator

Fit a cubic polynomial y = ax³ + bx² + cx + d to data using least squares. R², prediction, residuals, and comparison with linear and quadratic fits.

Statistics

Polynomial Regression Calculator

Fit polynomial of degree 1–6 to data. Compare models with R², adjusted R², AIC/BIC. Visualize fits and residuals.

Statistics

Quadratic Regression Calculator

Fit y = ax² + bx + c to data using least squares. Vertex, axis of symmetry, R², comparison with linear fit, residuals, F-test.

Statistics

Coefficient of Determination Calculator (R-squared)

Computes R², adjusted R², and related measures of fit from data or summary statistics. Explains how much variance in Y is explained by X.

Statistics

Linear Regression Calculator

Full simple linear regression: y = a + bx. Slope, intercept, R², standard errors, t-tests for coefficients, confidence and prediction intervals, ANOVA table.

Statistics

Absolute Uncertainty Calculator

Computes absolute uncertainty, relative (percentage) uncertainty, and propagates uncertainties through arithmetic operations. Essential for physics...

Statistics