6.8 Fitting Exponential Models to Data

Exponential and Logarithmic Models

Exponential and logarithmic models let you describe real-world patterns like population growth, radioactive decay, and market saturation using equations you can actually work with. Once you can fit a model to data, you can make predictions and quantify trends instead of just eyeballing a graph.

This section covers how to build exponential, logarithmic, and logistic models from data, and how to interpret what each one tells you.

Construction of Exponential Models

Exponential models follow the general form $y = ab^x$, where:

• $a$ is the initial value (the y-intercept, or the output when $x = 0$)
• $b$ is the growth or decay factor
  • If $b > 1$, you have exponential growth (e.g., population increase, compound interest)
  • If $0 < b < 1$, you have exponential decay (e.g., radioactive decay, depreciation)

To construct an exponential model from a data set:

1. Identify the initial value to determine $a$. This is the y-value when $x = 0$.
2. Calculate the growth/decay factor $b$ by dividing consecutive y-values: $b = \frac{y_2}{y_1}$ This ratio should be roughly constant across your data if an exponential model is a good fit. If $b$ comes out less than 1, that's decay; greater than 1, that's growth. (You don't flip the fraction for decay; the ratio naturally falls below 1.)
3. Write the model by substituting $a$ and $b$ into $y = ab^x$.

Quick example: Suppose a bacteria colony starts at 500 and triples every hour. Then $a = 500$, $b = 3$, and the model is $y = 500 \cdot 3^x$, where $x$ is time in hours.

Once you have the model, you can plug in future $x$-values to predict outcomes.

Development of Logarithmic Models

Logarithmic models follow the general form $y = a + b\ln(x)$, where:

• $a$ is the y-intercept (the value of $y$ when $\ln(x) = 0$, i.e., when $x = 1$)
• $b$ controls the rate of increase or decrease
  • $b > 0$: logarithmic growth (e.g., learning curves, where early gains are large but progress slows)
  • $b < 0$: logarithmic decrease (e.g., cooling rates)

These models are useful for situations with diminishing returns, where each additional unit of input produces a smaller and smaller change in output.

To develop a logarithmic model from data:

1. Transform the x-values by computing $\ln(x)$ for each data point.
2. Plot the transformed data as $(\ \ln(x),\ y\ )$ on a scatter plot. If a logarithmic model fits well, this plot should look roughly linear.
3. Find the line of best fit for the transformed data. The y-intercept of that line gives you $a$, and the slope gives you $b$.
4. Write the model by substituting $a$ and $b$ into $y = a + b\ln(x)$.

The key idea: taking the natural log of the x-values "straightens out" logarithmic data, turning a curve into a line you can analyze with linear techniques.

Interpretation of Logistic Growth Models

Logistic growth models handle situations where growth starts fast but levels off due to some limiting factor. The general form is:

$P(t) = \frac{L}{1 + ae^{-kt}}$

• $P(t)$ is the population (or quantity) at time $t$
• $L$ is the carrying capacity, the maximum value the population can reach given limited resources
• $a$ is a constant determined by the initial population
• $k$ is a positive constant that controls how quickly growth occurs

To build a logistic model:

1. Identify the carrying capacity $L$ from the context (e.g., maximum number of users a platform can support, or the resource limit for an animal population).

2. Determine the initial population $P_0$ at $t = 0$. Since $P(0) = \frac{L}{1 + a}$, you can solve for $a$: $a = \frac{L}{P_0} - 1$.

3. Use an additional data point to solve for $k$. Plug in a known $(t, P(t))$ pair and solve the equation for $k$.

Interpreting the model:

• Early on, the population grows nearly exponentially because it's far from the carrying capacity.
• At the inflection point, the population reaches $\frac{L}{2}$ (half the carrying capacity). This is where growth is fastest.
• As $t$ gets large, $P(t)$ approaches $L$ asymptotically. The curve flattens out because resources constrain further growth.

The S-shaped logistic curve shows up in population biology, disease spread, technology adoption, and market saturation.

Data Modeling and Analysis

When you have a data set and need to decide which model fits best, a few tools help:

• Regression analysis uses computational methods (often your calculator or software) to find the equation that best fits your data. Most graphing calculators have built-in exponential regression (ExpReg), logarithmic regression (LnReg), and logistic regression (Logistic) functions.
• Scatter plots are your first step. Plot the raw data and look at the shape. An upward-curving pattern suggests exponential growth; a curve that rises quickly then levels off suggests logarithmic or logistic behavior.
• Residuals measure the difference between each actual data point and the value your model predicts. Smaller residuals mean a better fit. If your residuals show a clear pattern (like curving up then down), your model type might be wrong, and you should try a different one.