📚 Advanced Study Notes on Averages
MIT‑Level Treatment – Aptitude & Advanced Mathematics
Goal:
- Develop a rigorous, first‑principles understanding of averages.
- Explore deeper theory, edge cases, and advanced inequalities.
- Present a challenging problem that tests mastery of the concepts.
1. Introduction
Averages (or means) are central in mathematics, statistics, and applied sciences.
They provide a single representative value summarizing a collection of data or a function’s behaviour.
Despite their ubiquity, many subtleties arise when we push into higher mathematics: the choice of mean, its properties, its continuity, and its role in optimisation and probability.
2. Basic Definitions (First Principles)
Let (x_1,\dots,x_n \in \mathbb{R}) be a finite multiset of real numbers.
Let (w_1,\dots,w_n \ge 0) be non‑negative weights with (\sum_{i=1}^n w_i = 1).
| Mean | Formula | Conditions | Interpretation |
|---|---|---|---|
| Arithmetic Mean (AM) | (\displaystyle A_n = \frac{1}{n}\sum_{i=1}^n x_i) | None | Simple average. |
| Weighted Arithmetic Mean (WAM) | (\displaystyle \bar{x}w = \sum{i=1}^n w_i x_i) | (\sum w_i = 1) | Average weighted by importance. |
| Geometric Mean (GM) | (\displaystyle G_n = \left(\prod_{i=1}^n x_i\right)^{1/n}) | (x_i > 0) | Multiplicative counterpart of AM. |
| Harmonic Mean (HM) | (\displaystyle H_n = \frac{n}{\displaystyle\sum_{i=1}^n \frac{1}{x_i}}) | (x_i > 0) | Useful for rates. |
| Root‑Mean‑Square (RMS) | (\displaystyle R_n = \sqrt{\frac{1}{n}\sum_{i=1}^n x_i^2}) | None | Quadratic average. |
| Generalised Mean (Power Mean) | (\displaystyle M_p = \left(\frac{1}{n}\sum_{i=1}^n x_i^p\right)^{1/p},; p \in \mathbb{R}\setminus{0}) | (x_i \ge 0) | Family of means parameterised by (p). |
| Logarithmic Mean (LM) | (\displaystyle L(a,b) = \frac{b-a}{\ln b - \ln a}) for (a\neq b) | (a,b>0) | Between geometric and arithmetic means. |
Remark: For (p=0) we define (M_0 = \lim_{p\to 0} M_p = \exp!\bigl(\frac{1}{n}\sum\ln x_i\bigr)) (the geometric mean).
3. Fundamental Properties (From First Principles)
3.1 Linearity of the Arithmetic Mean
For any scalars (a,b) and any real numbers (x_i):
[ A_n(a x_i + b) = a A_n(x_i) + b. ]
Proof:
[
A_n(a x_i + b) = \frac{1}{n}\sum (a x_i + b) = a\frac{1}{n}\sum x_i + b = a A_n(x_i) + b.
]
3.2 Symmetry
All unweighted means are symmetric: permuting the indices does not change the value.
3.3 Monotonicity (Ordering)
If (x_i \le y_i) for all (i), then (M_p(x) \le M_p(y)) for any (p>0).
Proof (sketch): Since (t\mapsto t^p) is increasing for (p>0), summing preserves inequality, taking the (p)-th root preserves it as well.
3.4 Homogeneity
For any scalar (c>0),
[ M_p(c x_i) = c, M_p(x_i). ]
Proof:
[
M_p(c x_i) = \Bigl(\frac{1}{n}\sum (c x_i)^p\Bigr)^{1/p} = c,\Bigl(\frac{1}{n}\sum x_i^p\Bigr)^{1/p} = c,M_p(x_i).
]
3.5 Weighted Means as Convex Combinations
A weighted arithmetic mean is a convex combination of the data points.
By convexity of the identity function, it lies between the minimum and maximum:
[ \min_i x_i \le \bar{x}_w \le \max_i x_i. ]
4. Classical Inequalities (Theoretical Backbone)
4.1 AM–GM Inequality
For (x_i>0),
[ A_n ;\ge; G_n,\qquad \text{