StatisticsCALCULATOR LAB
Statistics

Exponential Regression Calculator

Fit an exponential growth or decay model to paired x and y values. The calculator estimates y = a·e^(bx) by applying least squares to the natural logarithms of positive y observations.

01 / YOUR INPUTSIEEE 754 Double Precision

Fit an exponential regression model

Try an example:
Use a period for decimals, for example 1.5. Enter one X and one Y per row. For pasting, use one pair per line with columns separated by a space or tab.
Paired observations
XYActions
Up to 500 complete pairs. Every row needs both X and Y. Negative X values are allowed.
Paste X and Y pairs
One pair per line. Tabs or spaces separate columns; use the selected decimal separator. Applying replaces the current rows.
02 / YOUR RESULT

Your exponential regression model

Enter at least two valid x–y pairs with positive y values, or choose an example, then select Calculate.

How to use this calculator

  1. Enter paired x and y observations. Every y value must be greater than zero.
  2. Select Calculate to fit the exponential model to the logarithms of the y values.
  3. Review the fitted equation, growth or decay factor, log-scale R² and the curve plotted against your observations.

Exponential regression equation

The model is ŷ = a e^(bx), where a > 0. Taking natural logarithms gives ln(y) = ln(a) + bx. The calculator fits this straight-line relationship by least squares, then transforms the intercept back with a = e^(intercept). This minimizes squared differences on the log scale, not on the original y scale.

Growth and decay interpretation

For an increase of one unit in x, the fitted y value is multiplied by e^b. The corresponding percentage change is 100(e^b − 1)%. Positive b indicates growth, negative b indicates decay and b = 0 indicates a constant fitted level.

Worked example — exponential growth

This example uses sample values, not your current inputs.

For (x, y) pairs (0, 2), (1, 6), (2, 18) and (3, 54), the observations follow y = 2 × 3^x exactly. The equivalent natural-exponential form is ŷ = 2e^(1.0986122887x). Each one-unit increase in x multiplies y by 3, which is a 200% increase.

Frequently asked questions

What is exponential regression used for?

It models data that change approximately by a constant multiplicative factor for each one-unit increase in x, such as some growth and decay processes.

Why must all y values be positive?

This calculator fits a straight line to ln(y). Natural logarithms of zero or negative real values are not defined, so such observations cannot be included in this log-linear model.

Does the calculator use nonlinear least squares?

No. It uses ordinary least squares on ln(y), then transforms the fitted intercept back. Fitting squared errors on the original y scale is a different method and can produce different coefficients.

How do I interpret the exponent b?

The multiplicative factor for one x unit is e^b. The percentage change is 100(e^b − 1)%, which is not generally equal to 100b%.

What does R² on the log scale mean?

It describes how much variation in ln(y) is accounted for by the fitted linear relationship. It is not the same as R² calculated from residuals in the original y values.

Can I extrapolate outside my data?

You can evaluate the fitted equation outside the observed x range, but exponential predictions can grow or shrink quickly and may be unreliable without a justified model.

Scientific foundations and academic references

The mathematical foundations, computational method and limitations of this tool.

Foundational research · 1805/1806

A.-M. Legendre

A.-M. Legendre, Nouvelles méthodes pour la détermination des orbites des comètes (1805/1806 publication record)

Legendre's early published least-squares work provides the optimization principle behind linear regression. For this calculator, the principle is applied to the transformed response ln(y); this is a modern application rather than a claim that Legendre introduced log-linear exponential fitting.

Read source (opens in a new tab)

Theoretical development · 1809

C. F. Gauss

C. F. Gauss, Theoria motus corporum coelestium (1809)

Gauss's 1809 work advanced the analysis of fitting observations. Modern numerical linear algebra provides stable formulations of the linear least-squares step used after logarithmic transformation.

Read source (opens in a new tab)

Golub & Van Loan, Matrix Computations, 4th ed. (2013)

Read source (opens in a new tab)

Authoritative technical references

R stats, lm

R stats, lm

R's lm documentation specifies linear least-squares models. The same procedure can be applied to ln(y) as the response, provided all observed y values are positive. Its fitted coefficients are then converted to the original exponential equation.

Read source (opens in a new tab)

NumPy, Polynomial.fit (degree one)

NumPy, Polynomial.fit (degree one)

Read source (opens in a new tab)

Computational method and verification

Numerical computation

The estimator computes z = ln(y), fits z = α + bx by centered least squares and returns a = exp(α). All input y values must be strictly positive. The reported R² uses variation in log-transformed y, not variation in original y. Floating-point operations use IEEE 754 double precision; growth/decay predictions and exponentiation can overflow outside a reasonable domain. Display precision does not affect fitting. Residual standard error (log scale) = sqrt(SSElog / (n − 2)), available only when n > 2.

(0,2), (1,6), (2,18), (3,54)
a = 2; b = ln(3); factor = 3; change = +200%
(0,100), (1,50), (2,25)
a = 100; b = −ln(2); factor = 0.5; change = −50%
Constant positive Y, distinct X values
R² (log scale) undefined

Scope and limitations

The model ŷ = a e^(bx) is fitted by ordinary least squares after transforming the positive responses with z = ln(y). This minimizes squared errors in z, not in the original response y. The reported R² is calculated on the log scale and is undefined when log(y) has no variation. The growth factor for one x unit is e^b, and the corresponding percentage change is 100(e^b−1)%. Extreme exponentiation can overflow or underflow in double precision. Residual standard error (log scale) = sqrt(SSElog / (n − 2)), available only when n > 2.

Read full methodology

Cite this tool

Use the following citation when referencing this calculator.

APA 7th edition

Statistics Calculator Lab. (2026). Exponential Regression Calculator [Online calculator]. https://statscalculator.org/exponential-regression-calculator/

Show BibTeX
@misc{statscalculator_exponential_regression_2026,
  author = {{Statistics Calculator Lab}},
  title = {Exponential Regression Calculator},
  year = {2026},
  howpublished = {Online calculator},
  url = {https://statscalculator.org/exponential-regression-calculator/}
}