Why Rate and Scale Need Care
An exponential model describes a nonnegative quantity such as a waiting time or a hypothetical service duration. It is commonly written using a rate, but some sources instead specify a scale. These parameters are reciprocals, not interchangeable labels: using the scale where a formula requires the rate changes the model.
The parameter choice also carries units. If \(X\) is measured in hours, a rate is measured per hour, while a scale is measured in hours. Keeping those units visible makes it easier to catch a mistaken conversion and to interpret the density correctly. This tutorial establishes the exponential density and its parameterizations; probabilities and other properties of the model require additional results.
The support tells us which values the variable can take under the model. Negative waiting times are not allowed, so the density is zero below zero. At zero and for positive \(x\), the density follows the exponential formula. The value at a single endpoint does not change the integral of a density, but stating the support precisely prevents accidental use of the formula for negative values.
Rate and Scale Parameterizations
The rate parameter describes the scale of the exponential decrease in the density. A larger rate makes the density decrease more quickly as \(x\) increases. The alternative scale parameter, often written \(\theta\), is defined by \(\theta=1/\lambda\). It has the same units as \(X\), and expressing the same density in terms of \(\theta\) gives the following form.
The two formulas describe the same family of distributions when the parameters are reciprocals. If a problem says that the scale is 5 days, for example, then the rate is \(1/5\) per day. If it gives a rate of \(0.2\) per day, the scale is 5 days. A parameter’s numerical value alone is not enough to identify its meaning; look for whether the problem supplies a rate or a scale and check the units.
| Parameterization | Parameter meaning | Units when \(X\) is in hours | Density form for \(x\geq 0\) |
|---|---|---|---|
| Rate | \(\lambda\) | per hour | \(\lambda e^{-\lambda x}\) |
| Scale | \(\theta=1/\lambda\) | hours | \((1/\theta)e^{-x/\theta}\) |
Some references use different symbols for the scale or rate. The reliable approach is to read the stated definition rather than infer a parameter’s role from its symbol. In this course, \(\lambda\) denotes the rate and \(\theta\) denotes its reciprocal scale. The units provide a second check: a quantity added to the exponent must be dimensionless, so \(\lambda x\) and \(x/\theta\) have no units.
Why the Formula Is a Valid Density
A continuous density must be nonnegative and have total integral 1 over its support. The exponential formula is nonnegative for \(\lambda>0\). Its total integral verifies the normalization requirement.
For the evaluation, \(\lambda>0\) ensures that \(e^{-\lambda x}\) tends to zero as \(x\) tends to infinity; at \(x=0\), it equals 1. Therefore the formula integrates to 1 for every positive rate. Equivalently, substituting \(\theta=1/\lambda\) into the scale form gives the same integral. If the parameter is zero or negative, this argument fails and the formula does not define an exponential density.
Worked Examples
Worked Example: Convert a Scale and Evaluate the Density
A hypothetical equipment-monitoring model represents a nonnegative operating time \(X\) in days with an exponential distribution. The scale is 4 days. Find the rate, write the density using both parameterizations, and evaluate the density at 6 days. Report the density to four decimal places.
Method. Convert the scale to a rate using \(\lambda=1/\theta\), then substitute the same parameter values into the rate and scale forms of the density.
Step 1: Find the rate. The scale is \(\theta=4\) days, so $$ \lambda=\frac{1}{\theta}=\frac{1}{4}=0.25\text{ per day}. $$ The rate is positive, as required.
Step 2: Write the density. For \(x\geq 0\), the rate form and the scale form are $$ f_X(x)=0.25e^{-0.25x} \qquad\text{and}\qquad f_X(x)=\frac{1}{4}e^{-x/4}. $$ Both expressions use \(x\) in days and give the same density.
Step 3: Evaluate at 6 days. $$ f_X(6)=0.25e^{-0.25(6)} =0.25e^{-1.5} =0.25(0.2231302) =0.0557826\text{ per day}. $$ Rounded to four decimal places, the density is \(0.0558\) per day.
Check and trap avoided. The rate and scale are reciprocals: \(0.25\times4=1\). The exponent uses the dimensionless product \((0.25\text{ per day})(6\text{ days})=1.5\). A common mistake is to use 4 as the rate, giving \(4e^{-24}\); that incorrectly treats a scale measured in days as a rate measured per day. Also, \(0.0558\) per day is a density value, not the probability that \(X\) equals 6 days.
Worked Example: Re-express a Rate in Different Time Units
A hypothetical call-routing model assigns an exponential waiting time \(X\) a rate of \(0.08\) per month. Express the model using years instead of months, find its scale in both units, and evaluate the density at 3 months in each representation. Report density values to four decimal places.
Method. Convert the rate and the waiting time together so that the exponent is unchanged. Since one year is 12 months, the numerical rate per year is 12 times the rate per month.
Step 1: Find the scale in months. $$ \theta_{\text{months}}=\frac{1}{0.08}=12.5\text{ months}. $$
Step 2: Convert the rate and scale to years. $$ \lambda_{\text{years}}=12(0.08)=0.96\text{ per year}, \qquad \theta_{\text{years}}=\frac{1}{0.96} =1.0416667\text{ years}. $$ The same scale expressed in months is \(12(1.0416667)=12.5\) months, agreeing with the first calculation.
Step 3: Evaluate the density at 3 months. In month units, \(x=3\) and \(\lambda=0.08\): $$ f_{X,\text{months}}(3) =0.08e^{-0.08(3)} =0.08e^{-0.24} =0.08(0.7866279) =0.0629302\text{ per month}. $$ In year units, 3 months is \(x=3/12=0.25\) years: $$ f_{X,\text{years}}(0.25) =0.96e^{-0.96(0.25)} =0.96e^{-0.24} =0.7551628\text{ per year}. $$
Step 4: Check the density conversion. There are 12 months in a year, so a density per year is 12 times the corresponding density per month: $$ 12(0.0629302)=0.7551624\text{ per year}, $$ which agrees with \(0.7551628\) per year apart from rounding in the intermediate values. To four decimal places, the values are \(0.0629\) per month and \(0.7552\) per year.
Check and trap avoided. The exponent is \(-0.24\) in both units, confirming that the point has been converted consistently. A common error is to divide the rate by 12 when converting a monthly rate to a yearly rate. That reverses the conversion: \(0.08\) per month corresponds to \(0.96\) per year, not \(0.0067\) per year.
Worked Example: Normalize a Proposed Exponential Form
A hypothetical model proposes a density of the form \(g(x)=c e^{-0.2x}\) for \(x\geq 0\), and \(g(x)=0\) for \(x<0\). Here \(x\) is measured in days. Determine the constant \(c\) that makes \(g\) a density, then identify the rate, scale, and density at 2 days. Report the density to four decimal places.
Method. Require the integral over the support to equal 1. Then compare the normalized formula with the exponential rate form.
Step 1: Normalize the proposed function. $$ \int_0^\infty c e^{-0.2x}\,dx =c\left[-\frac{1}{0.2}e^{-0.2x}\right]_0^\infty =c(5) =5c. $$ For a density, \(5c=1\), so \(c=0.2\). This is positive, and the function is nonnegative on its support.
Step 2: Identify the parameters. The normalized density is $$ g(x)=0.2e^{-0.2x},\qquad x\geq 0. $$ Comparing it with \(\lambda e^{-\lambda x}\) gives \(\lambda=0.2\) per day. The scale is $$ \theta=\frac{1}{0.2}=5\text{ days}. $$
Step 3: Evaluate the density at 2 days. $$ g(2)=0.2e^{-0.2(2)} =0.2e^{-0.4} =0.2(0.6703200) =0.1340640\text{ per day}. $$ Rounded to four decimal places, \(g(2)=0.1341\) per day.
Check and trap avoided. The normalized density integrates to \(5(0.2)=1\), and the rate and scale satisfy \(0.2(5)=1\). A common mistake is to set \(c=0.2\) by copying the exponent coefficient without checking normalization. In this example that happens to produce the right constant, but it is the integral condition—not a general rule that the multiplier equals the exponent coefficient—that establishes the answer.
Density Values, Units, and Model Checks
A density is not itself a probability. For a continuous random variable, probabilities are obtained by integrating the density over sets of values; the probability at any single exact point is zero. Consequently, a density value can exceed 1 without contradiction, particularly when the variable is measured in small units. Its units are the reciprocal of the units used for \(X\), such as per hour or per day.
Changing the unit changes the numerical density value, even though the underlying model has not changed. In the time-conversion example, the density at the same physical time had different numerical values per month and per year. The values agree after accounting for the conversion between the units. This is why reporting a density without its measurement unit can be misleading.
Before using an exponential density, check the model and its parameters in a deliberate order:
Confirm that the modeled variable is nonnegative, and note whether it is measured in hours, days, or another unit.
Determine whether the supplied value is a rate or a scale. Do not decide from the numerical value alone.
Verify that \(\lambda>0\), \(\theta>0\), and \(\lambda\theta=1\), with rate and scale units reciprocal to each other.
Substitute a rate into \(\lambda e^{-\lambda x}\), or a scale into \((1/\theta)e^{-x/\theta}\), keeping \(x\) in the same units as the scale.
The model does not apply as written if the variable can take negative values or if the parameter is not positive. In those cases, the stated support or density is inconsistent with the proposed model. A density with an additional constant multiplier also needs a normalization check rather than an assumption that it is already exponential.
Check Your Understanding
For each problem, identify the parameterization before calculating, and keep the stated units attached to rates, scales, and density values.
- A hypothetical sensor’s waiting time is exponential with scale 8 minutes. Find the rate and state its units. (Answer: \(\lambda=0.125\) per minute.)
- A service model uses rate \(0.5\) per hour. Find the scale in hours and write the density on its support. (Answer: \(\theta=2\) hours; \(f_X(x)=0.5e^{-0.5x}\) for \(x\geq0\), and \(0\) for \(x<0\).)
- A hypothetical device lifetime in days has rate \(0.4\) per day. Evaluate its density at 3 days to four decimal places. (Answer: \(f_X(3)=0.4e^{-1.2}=0.1205\) per day.)
- A monitoring model specifies rate \(0.03\) per second. Convert the rate to per minute and find the scale in minutes. (Answer: \(\lambda=1.8\) per minute and \(\theta=1/1.8=0.5556\) minutes.)
- A proposed density is \(g(x)=c e^{-0.4x}\) for \(x\geq0\), and zero otherwise, with \(x\) measured in hours. Find \(c\), then give the rate and scale with units. (Answer: \(c=0.4\), \(\lambda=0.4\) per hour, and \(\theta=2.5\) hours.)