How the Beta Shape Parameters Shape a Density
A beta random variable is supported on the interval from 0 to 1, but its density can have very different shapes within that interval. It can be symmetric and mound-shaped, concentrated near one endpoint, or U-shaped with high density near both endpoints. The two positive shape parameters determine these features. Being able to read them correctly is useful whenever a model uses a beta distribution for a quantity constrained to lie between 0 and 1.
The powers in the density give a first indication of its endpoint behavior. As \(x\) approaches 0, the factor \(x^{\alpha-1}\) tends to infinity when \(\alpha<1\), remains equal to 1 when \(\alpha=1\), and tends to 0 when \(\alpha>1\). The corresponding behavior near 1 is determined by \(\beta\), through \((1-x)^{\beta-1}\). These limits describe the density, not point probabilities: a continuous random variable still has probability 0 at either single endpoint.
The mean and variance, established in the earlier material on beta distribution moments, are \(E[X]=\alpha/(\alpha+\beta)\) and \(\operatorname{Var}(X)=\alpha\beta/((\alpha+\beta)^2(\alpha+\beta+1))\). The mean describes a center of balance, not necessarily the location of the density’s highest point. That highest point is the mode, and its location depends on which parameter ranges apply.
Symmetry and the Direction of Skew
When \(\alpha=\beta\), exchanging \(x\) and \(1-x\) leaves the density unchanged. The distribution is symmetric around \(1/2\). When the parameters differ, the density is generally asymmetric. The standardized third central moment, or skewness, has the sign of \(\beta-\alpha\):
Thus, if \(\alpha<\beta\), the skewness is positive: the density tends to extend farther toward the right, even though more of its mass is toward the left. If \(\alpha>\beta\), the skewness is negative and the longer tail is toward the left. The formula gives the direction of skew; it does not, by itself, describe every visual feature of the density.
Equal parameters do not always mean a mound-shaped density. For example, \(\operatorname{Beta}(4,4)\) is symmetric and has a central peak, while \(\operatorname{Beta}(0.6,0.6)\) is symmetric and has high density near both endpoints. The size of the parameters, as well as whether they are equal, matters.
Finding the Mode
For a density that has a finite interior maximum, differentiate its logarithm. This avoids carrying the constant \(1/B(\alpha,\beta)\) through the derivative. When both parameters exceed 1, the density is 0 at both endpoints and has an interior maximum.
For \(0<x<1\), \(\log f_X(x)=(\alpha-1)\log x+(\beta-1)\log(1-x)-\log B(\alpha,\beta)\).
\(\frac{d}{dx}\log f_X(x)=\frac{\alpha-1}{x}-\frac{\beta-1}{1-x}=0\).
Multiplying by \(x(1-x)\) gives \((\alpha-1)(1-x)-(\beta-1)x=0\), so \(x=(\alpha-1)/(\alpha+\beta-2)\).
The second derivative of the log density is \(-(\alpha-1)/x^2-(\beta-1)/(1-x)^2\), which is negative when both parameters exceed 1. The stationary point is therefore the mode.
The endpoint cases require care. If \(\alpha=1\) and \(\beta>1\), the density decreases from its maximum at 0. If \(\alpha>1\) and \(\beta=1\), it increases to its maximum at 1. If both parameters are 1, the density is flat. When a parameter is below 1, the density is unbounded at its corresponding endpoint; it is often described as having a boundary mode there, meaning that the density grows without bound as it approaches that endpoint.
| Parameter conditions | Density shape and mode behavior |
|---|---|
| \(\alpha>1,\ \beta>1\) | One interior mode at \((\alpha-1)/(\alpha+\beta-2)\). |
| \(\alpha=1,\ \beta=1\) | Uniform density; every point in \([0,1]\) is a mode. |
| \(\alpha=1,\ \beta>1\) | Decreases from a finite maximum at 0. |
| \(\alpha>1,\ \beta=1\) | Increases to a finite maximum at 1. |
| \(\alpha<1,\ \beta\geq 1\) | Density is unbounded near 0; boundary concentration at 0. |
| \(\alpha\geq 1,\ \beta<1\) | Density is unbounded near 1; boundary concentration at 1. |
| \(\alpha<1,\ \beta<1\) | Density is unbounded near both endpoints and has a U-shape. |
In the last three rows, “boundary mode” is a useful shape description, but the density has no finite maximum value at the endpoint where it diverges. When both parameters are below 1, the density has a stationary point inside the interval, but it is a minimum, not a maximum. Applying the interior-mode formula without checking the parameter conditions can therefore give the wrong interpretation.
Worked Examples
Worked Example: An Interior Mode and a Skew Direction
A hypothetical model specifies a proportion \(X\) by \(X\sim\operatorname{Beta}(4,7)\). Determine whether the density is symmetric, find its mode and mean, and identify the direction of skewness.
Method. Compare the parameters for symmetry, use the interior-mode formula only after checking its conditions, and use the sign of the skewness formula for the tail direction.
Both parameters exceed 1, so the mode is interior. They are unequal, so the density is not symmetric. The mode is \(\frac{4-1}{4+7-2}=\frac{3}{9}=\frac{1}{3}\).
The mean is \(E[X]=\frac{4}{4+7}=\frac{4}{11}\approx0.3636\). The mode is approximately \(0.3333\), which is less than the mean.
Since \(\beta-\alpha=7-4=3>0\), the skewness is positive. Its value is $$ \gamma_1= \frac{2(7-4)\sqrt{4+7+1}} {(4+7+2)\sqrt{4\cdot7}} =\frac{6\sqrt{12}}{13\sqrt{28}} \approx 0.3021. $$ The density has a right-skewed shape. Check: the calculated mode lies between 0 and 1, as required, and the positive skewness agrees with \(\beta>\alpha\). The common mistake is to call the distribution symmetric because both parameters are positive; symmetry requires that they be equal.
Worked Example: A Symmetric U-Shaped Density
A hypothetical model assigns \(X\sim\operatorname{Beta}(0.5,0.5)\) to a proportion. Classify its shape, determine its center and endpoint behavior, and decide whether the interior-mode formula can be used.
Method. First compare the parameters, then check whether they are above or below 1 before interpreting the mode.
The parameters are equal, so the distribution is symmetric about \(1/2\). Its mean is \(E[X]=0.5/(0.5+0.5)=0.5\). Both parameters are less than 1. Therefore \(x^{-0.5}\) causes the density to diverge as \(x\) approaches 0, and \((1-x)^{-0.5}\) causes it to diverge as \(x\) approaches 1. The density is U-shaped, with unbounded density near both endpoints.
The interior-mode formula requires both parameters to exceed 1 and cannot be used here. Substitution would give \(\frac{0.5-1}{0.5+0.5-2}=\frac{-0.5}{-1}=0.5\), but this stationary point is a minimum, not a mode. Check: the symmetric density has equal endpoint behavior and a central minimum, consistent with a U-shape. The mean \(0.5\) does not imply a central peak.
Worked Example: One Finite Boundary Maximum and One Singular Boundary
For a hypothetical proportion, let \(X\sim\operatorname{Beta}(1,4)\). Find the density, classify its mode and skewness, and compare the mode with the mean.
Method. Use the beta density, then classify the endpoint case rather than applying the interior-mode formula.
Here \(B(1,4)=\Gamma(1)\Gamma(4)/\Gamma(5)=1\cdot6/24=1/4\). Thus, for \(0<x<1\), \(f_X(x)=x^0(1-x)^3/(1/4)=4(1-x)^3\). This density decreases as \(x\) increases, so its finite maximum on the interval is at \(x=0\), with density value 4. The parameter case \(\alpha=1,\ \beta>1\) agrees with that decreasing shape.
The mean is \(E[X]=1/(1+4)=0.2\). Since \(\beta-\alpha=3>0\), the distribution is positively skewed. The mode is 0, which is less than the mean. Check: \(4(1-0)^3=4\) and the density tends to 0 as \(x\) approaches 1. A common error is to substitute into the interior-mode formula, whose denominator here is \(1+4-2=3\), producing 0. That happens to identify the boundary location, but the formula’s interior-maximum conditions are not satisfied.
Interpreting Shape Without Confusing Mean and Mode
The mode and mean answer different questions. The mode locates the highest point of a density when a finite maximum exists, while the mean is its probability-weighted average. For \(\alpha,\beta>1\), their difference is $$ \frac{\alpha-1}{\alpha+\beta-2}-\frac{\alpha}{\alpha+\beta} =\frac{\alpha-\beta}{(\alpha+\beta-2)(\alpha+\beta)}. $$ Thus, if \(\alpha>\beta\), the interior mode is to the right of the mean; if \(\alpha<\beta\), it is to the left. This is consistent with negative skew in the first case and positive skew in the second.
A reliable classification proceeds in a fixed order:
- Compare \(\alpha\) and \(\beta\) to assess symmetry and the sign of skewness.
- Check each parameter against 1 to determine how the density behaves at its endpoint.
- Use the interior-mode formula only when both parameters exceed 1.
- For boundary cases, describe whether the density has a finite endpoint maximum or becomes unbounded near the endpoint.
These checks matter because a single parameter change can change the type of mode. For instance, a parameter just above 1 gives density tending to 0 at its endpoint, while a parameter just below 1 gives density that grows without bound there. A numerical mode formula should never replace checking the shape conditions first.
Check Your Understanding
For each model, classify the shape and use the parameters to check your conclusion.
- A hypothetical variable has distribution \(\operatorname{Beta}(6,3)\). Find its mean and mode, and state whether its skewness is positive or negative. (Answer: Mean \(=2/3\); mode \(=5/7\); negative skewness.)
- A hypothetical proportion has distribution \(\operatorname{Beta}(1,5)\). State whether its density is increasing or decreasing, locate its mode, and find its mean. (Answer: Decreasing; finite boundary mode at 0; mean \(=1/6\).)
- A model uses \(\operatorname{Beta}(0.7,0.7)\). Classify the symmetry and endpoint behavior, and state the mean. (Answer: Symmetric U-shape, unbounded near both endpoints; mean \(=1/2\).)
- For \(X\sim\operatorname{Beta}(3,8)\), calculate the mode and mean, then state the skew direction and which of the mode or mean is larger. (Answer: Mode \(=2/9\); mean \(=3/11\); positive skewness; the mean is larger.)
- Compare hypothetical models \(\operatorname{Beta}(2,1)\) and \(\operatorname{Beta}(1,2)\). Give the direction of density change, the mode, and the mean for each. (Answer: \(\operatorname{Beta}(2,1)\) increases to mode 1 and has mean \(2/3\); \(\operatorname{Beta}(1,2)\) decreases from mode 0 and has mean \(1/3\).)