nonweighted

Nonweighted describes methods, analyses, or structures that do not assign weights to observations, variables, edges, or components. It contrasts with weighted approaches, which attach a numerical weight to reflect importance, frequency, reliability, or design features.

In statistics and data analysis, nonweighted commonly refers to equal weighting. A nonweighted mean uses all observations with equal influence. Nonweighted regression (ordinary least squares) assigns equal influence to all data points, whereas weighted least squares adjusts for varying variance or design. In survey sampling, nonweighted estimates treat each sampled unit equally, in contrast to weighted estimates that compensate for unequal selection probability, nonresponse, or population composition.

In graph theory and network analysis, a nonweighted or unweighted graph has edges without weights, so all connections are considered equally strong. Analyses on unweighted graphs often use simple adjacency relationships and unweighted distance measures; when weights exist, results may differ significantly. In machine learning, some models or features may be treated as nonweighted, for example in simple classifiers where each feature or vote counts equally.

Limitations: Nonweighted methods can bias results if the data have unequal reliability, sampling probabilities, or heterogeneity. Weighted methods aim to correct such distortions. Practitioners choose nonweighted approaches for simplicity, interpretability, or when weights are inappropriate or unavailable.

See also

- Weight (statistics)

- Weighted average

- Ordinary least squares

- Unweighted graph