Bimodal Distribution
Statistics & ProbabilityA bimodal distribution has two distinct peaks, indicating that data clusters around two different values.
Definition
A bimodal distribution is a dataset with two different peaks or clusters. Instead of one most-common value, there are two groups that appear most often.
Example
Survey scores from a class of students where half loved the lesson and half disliked it might cluster around $2/10$ and $9/10$, creating two peaks in the distribution.
Key Insight
Two peaks often mean two different groups are mixed together. Separating the groups and analyzing them individually usually makes more sense than reporting one overall average.
Definition
A bimodal distribution has two distinct local maxima (peaks) in its frequency or probability distribution. It often indicates that the data comes from two different subpopulations that have been combined. The overall mean falls between the two peaks and may not represent either group well.
Example
Heights of a mixed group of adults and children will be bimodal: one peak around adult average heights (5'7") and one around children's heights (4'2"). The overall mean does not represent either group.
Key Insight
Bimodal distributions are a sign to look for hidden structure: are two groups, time periods, or conditions mixed together? Clustering algorithms can help identify and separate the two modes.
Definition
A bimodal distribution has two modes, which formally means its PDF has two local maxima. Mixture models represent bimodal distributions as weighted sums of component distributions: $f(x) = \pi f_1(x) + (1-\pi) f_2(x)$, where $f_1$ and $f_2$ are typically normal PDFs and $\pi$ is the mixing proportion. Parameters are estimated via the EM (Expectation-Maximization) algorithm.
Example
A Gaussian mixture model with two components $N(\mu_1=3, \sigma_1=1)$ and $N(\mu_2=8, \sigma_2=1.5)$ with mixing weight $\pi=0.4$ has modes at approximately $3$ and $8$. The EM algorithm iterates between assigning soft memberships and updating parameters until convergence.
Key Insight
The Hartigan dip test is a formal nonparametric test for unimodality vs. multimodality. Silverman's bandwidth test uses KDE to test for a single mode. Both serve as diagnostic tools before applying mixture models.