Skip to content

Probability Theory Formula Sheet and Short Notes

One-page CSIR NET revision of descriptive statistics and probability: moments, regression, Bayes, CDFs, MGFs, characteristic functions, convergence, LLN, CLT.

  • 10 min read
  • Updated
On this page · 7 sections
  1. The paper in one table
  2. Descriptive statistics and EDA
  3. Probability spaces, Bayes, independence
  4. Random variables and moments
  5. Joint, conditional, characteristic functions
  6. Inequalities and limit theorems
  7. Last-week checklist

Every formula, table and counter-example from Descriptive Statistics and Exploratory Data Analysis, Sample Space, Independence and Bayes’ Theorem, Random Variables, Distribution Functions and Moments, Joint, Marginal and Conditional Distributions; Characteristic Functions and Inequalities, Convergence and Limit Theorems, on one page.

The paper in one table

Per the NTA information bulletin for December 2026 (Mathematical Sciences, code 704):

PartGivenAttemptCorrectWrongMaximum
B4025+3−0.7575
C6020+4.75095

Part C: one or more options may be correct; marks only if you identify all correct options and mark no wrong one. No partial credit, no negative marking. Details on the chapter overview and the Mathematical Sciences overview.

Descriptive statistics and EDA

IdeaFact
Minimisers\( \sum (x_i - a)^2 \) at \( \bar{x} \); \( \sum \lvert x_i - a \rvert \) at the median
Identity\( \sum (x_i - a)^2 = \sum (x_i - \bar{x})^2 + n(\bar{x} - a)^2 \)
Means\( \bar{x} \ge G \ge H \) (positive data)
Coding \( u = (x - a)/h \)\( \bar{x} = a + h\bar{u} \); \( s_x^2 = h^2 s_u^2 \); \( \mu_r \) scales by \( h^r \)
Pooled variance\( \frac{n_1(s_1^2 + d_1^2) + n_2(s_2^2 + d_2^2)}{n_1 + n_2} \), \( d_i = \bar{x}_i - \bar{x} \)
Bounds on \( s \)\( s \le \frac{\text{range}}{2} \); MD about mean \( \le s \)
Central from raw (\( m = \mu_1' \))\( \mu_2 = \mu_2' - m^2 \); \( \mu_3 = \mu_3' - 3\mu_2'm + 2m^3 \); \( \mu_4 = \mu_4' - 4\mu_3'm + 6\mu_2'm^2 - 3m^4 \)
Shape\( \beta_1 = \frac{\mu_3^2}{\mu_2^3} \), \( \beta_2 = \frac{\mu_4}{\mu_2^2} \), \( \beta_2 \ge \beta_1 + 1 \ge 1 \)
Kurtosis anchorsuniform 1.8, normal 3, Laplace 6, exponential 9
Bowley skewness\( \frac{Q_3 + Q_1 - 2Q_2}{Q_3 - Q_1} \in [-1, 1] \)
Sheppardgrouped variance \( - \frac{h^2}{12} \)
Tukey fences\( Q_1 - 1.5\,\text{IQR} \), \( Q_3 + 1.5\,\text{IQR} \); outer at 3 IQR
Resistantmedian, IQR, MAD, trimmed mean (not mean, \( s \), range, \( r \))

Correlation and regression

$$b_{yx} = r\frac{s_y}{s_x}, \qquad b_{xy} = r\frac{s_x}{s_y}, \qquad b_{yx}b_{xy} = r^2, \qquad \tan\theta = \frac{1 - r^2}{\lvert r \rvert}\cdot\frac{s_xs_y}{s_x^2 + s_y^2}$$
  • Both lines pass through \( (\bar{x}, \bar{y}) \); slopes share the sign of \( r \); product \( > 1 \) means you swapped the lines.
  • \( r \) is free of origin and scale (sign flips if exactly one scale factor is negative); slopes are free of origin only: \( b_{vu} = \frac{h}{k}b_{yx} \).
  • \( r = 0 \): lines perpendicular (\( x = \bar{x} \), \( y = \bar{y} \)); \( r = \pm 1 \): lines coincide.
  • Residual variance \( s_y^2(1 - r^2) \).

Probability spaces, Bayes, independence

IdeaFact
\( \sigma \)-fields on an \( n \)-setBell number: 1, 2, 5, 15, 52
Inclusion–exclusion\( S_1 - S_2 + S_3 - \dots \); truncations alternate upper and lower bounds
Exactly \( m \) of \( n \) events\( \sum_{k \ge m} (-1)^{k-m}\binom{k}{m}S_k \)
Matching\( P(\text{no match}) = \sum_{k=0}^{n}\frac{(-1)^k}{k!} \to \frac{1}{e} \); \( n = 4 \): \( \frac{3}{8} \)
Birthdaybelow \( \frac{1}{2} \) at 23 people
Geometric\( P(\lvert X - Y \rvert < a) = 1 - (1 - a)^2 \); broken stick \( \frac{1}{4} \); Buffon \( \frac{2l}{\pi d} \)
Total probability\( P(A) = \sum P(A \mid B_i)P(B_i) \)
Bayes\( P(B_j \mid A) = \frac{P(A \mid B_j)P(B_j)}{\sum_i P(A \mid B_i)P(B_i)} \); posterior odds = likelihood ratio \( \times \) prior odds
Screening (1%, 95%, 5%)\( P(D \mid +) \approx 0.161 \)
Pólya urnevery draw has \( P(\text{red}) = \frac{r}{r+b} \); draws exchangeable, not independent
Mutual independence of \( n \) events\( 2^n - n - 1 \) product equations

Random variables and moments

IdeaFact
CDF iffnon-decreasing, right-continuous, \( F(-\infty) = 0 \), \( F(\infty) = 1 \)
Atom\( P(X = a) = F(a) - F(a-) \); countably many atoms at most
Intervals\( P(a < X \le b) = F(b) - F(a) \); \( P(a \le X < b) = F(b-) - F(a-) \)
Typesdiscrete + absolutely continuous + singular (Lebesgue decomposition)
Continuous vs densityno atoms vs absolutely continuous (Cantor: continuous, no density)
Inverse transform\( F^{-1}(U) \sim F \); \( F(X) \sim U(0, 1) \) iff \( F \) continuous
Monotone transform\( f_Y(y) = f_X(h(y))\lvert h'(y) \rvert \)
\( Y = X^2 \)\( f_Y(y) = \frac{f_X(\sqrt{y}) + f_X(-\sqrt{y})}{2\sqrt{y}} \)
Mean existsiff \( E\lvert X \rvert < \infty \)
Tail formula\( E X = \int_0^\infty P(X > x)\,dx \) (\( X \ge 0 \)); \( \sum_{n \ge 1} P(X \ge n) \) on \( \{0, 1, \dots\} \)
Momentsfinite \( s \)-th \( \Rightarrow \) all lower; tails \( \sim \lvert x \rvert^{-(p+1)} \): order \( < p \) only; \( t_k \): order \( < k \)
PGF\( G'(1) = E X \); \( \text{Var} = G''(1) + G'(1) - G'(1)^2 \)
MGFfinite near 0 \( \Rightarrow \) determines law, \( M^{(r)}(0) = E X^r \); \( M_{aX+b}(t) = e^{bt}M(at) \)
No MGFCauchy (no mean); lognormal (all moments, \( M(t) = \infty \) for \( t > 0 \))
Memorylessexponential (continuous), geometric (discrete)
LawMeanVarianceMGF
Binomial(\( n, p \))\( np \)\( npq \)\( (q + pe^t)^n \)
Poisson(\( \lambda \))\( \lambda \)\( \lambda \)\( e^{\lambda(e^t - 1)} \)
Geometric on \( \{1, 2, \dots\} \)\( \frac{1}{p} \)\( \frac{q}{p^2} \)\( \frac{pe^t}{1 - qe^t} \)
Uniform(\( a, b \))\( \frac{a+b}{2} \)\( \frac{(b-a)^2}{12} \)\( \frac{e^{bt} - e^{at}}{t(b-a)} \)
Exponential (rate \( \lambda \))\( \frac{1}{\lambda} \)\( \frac{1}{\lambda^2} \)\( \frac{\lambda}{\lambda - t} \)
Gamma(\( \alpha \), rate \( \lambda \))\( \frac{\alpha}{\lambda} \)\( \frac{\alpha}{\lambda^2} \)\( (\frac{\lambda}{\lambda - t})^\alpha \)
Normal(\( \mu, \sigma^2 \))\( \mu \)\( \sigma^2 \)\( e^{\mu t + \sigma^2t^2/2} \)

Joint, conditional, characteristic functions

IdeaFact
Joint CDFneeds the rectangle inequality, not just monotonicity
Independencedensity factors on a product support
Uncorrelated vs independentsame only for jointly normal; \( (X, X^2) \), \( (X, SX) \) are counter-examples
Tower\( E[E(Y \mid X)] = E Y \)
Total variance\( \text{Var}\,Y = E\,\text{Var}(Y \mid X) + \text{Var}\,E(Y \mid X) \)
Random sum\( E S = E N\,E X \); \( \text{Var}\,S = E N\,\text{Var}\,X + \text{Var}\,N\,(E X)^2 \)
Best predictor\( E(Y \mid X) \); best linear: regression line
Bivariate normal\( Y \mid X = x \sim N(\mu_2 + \rho\frac{\sigma_2}{\sigma_1}(x - \mu_1), \sigma_2^2(1 - \rho^2)) \); \( \rho = 0 \iff \) independent
Poisson\( X \mid X + Y = n \sim \text{Bin}(n, \frac{\lambda}{\lambda + \mu}) \)
Jacobian\( f_{U,V} = f_{X,Y}\lvert \partial(x, y)/\partial(u, v) \rvert \)
Gamma–beta\( X + Y \sim \text{Gamma}(\alpha + \beta) \) independent of \( \frac{X}{X+Y} \sim \text{Beta}(\alpha, \beta) \)
Exponentials\( \min \) of i.i.d. rate \( \lambda \): rate \( 2\lambda \)
Normals\( X/Y \) Cauchy; \( X^2 + Y^2 \) exponential with mean 2

Characteristic functions: \( \varphi(0) = 1 \), \( \lvert \varphi \rvert \le 1 \), \( \varphi(-t) = \overline{\varphi(t)} \), uniformly continuous, real iff symmetric, \( \varphi_{aX+b}(t) = e^{ibt}\varphi(at) \), products for independent sums, \( \varphi^{(k)}(0) = i^kE X^k \), uniqueness, inversion, Lévy continuity.

LawCFLawCF
\( \pm 1 \)\( \cos t \)\( U(-a, a) \)\( \frac{\sin at}{at} \)
Binomial\( (q + pe^{it})^n \)Exponential\( \frac{\lambda}{\lambda - it} \)
Poisson\( e^{\lambda(e^{it} - 1)} \)Laplace\( \frac{1}{1 + t^2} \)
Normal\( e^{i\mu t - \sigma^2t^2/2} \)Cauchy\( e^{-\lvert t \rvert} \)

CFs: \( e^{-\lvert t \rvert^\alpha} \) (\( 0 < \alpha \le 2 \)), \( \max(1 - \lvert t \rvert, 0) \), \( \lvert \varphi \rvert^2 \), \( \text{Re}\,\varphi \), \( \varphi^n \). Not CFs: \( e^{-t^4} \), \( \frac{1}{1 + t^4} \) (flat at 0, so degenerate), \( \sin t \), anything with modulus above 1. \( \varphi(t)^2 \ne \varphi_{2X}(t) = \varphi(2t) \).

Inequalities and limit theorems

InequalityStatement
Markov\( P(X \ge a) \le \frac{E X}{a} \), \( X \ge 0 \)
Chebyshev\( P(\lvert X - \mu \rvert \ge k\sigma) \le \frac{1}{k^2} \); sharp
Cantelli\( P(X - \mu \ge \lambda) \le \frac{\sigma^2}{\sigma^2 + \lambda^2} \)
Chernoff\( P(X \ge a) \le \inf_{t > 0} e^{-ta}M(t) \)
Jensenconvex \( g \): \( g(E X) \le E g(X) \); \( E\frac{1}{X} \ge \frac{1}{E X} \); \( E\log X \le \log E X \)
Lyapunov\( (E\lvert X \rvert^r)^{1/r} \) non-decreasing in \( r \)
graph LR
    AS[a.s.] --> P[probability]
    LR[L^r] --> P
    P --> D[distribution]
    D -.->|constant limit| P
    P -.->|subsequence| AS
Sequencea.s.P\( L^1 \)d
\( n\mathbf{1}_{(0, 1/n)} \)yesyesnoyes
typewriter indicatorsnoyesyesyes
\( -X \) for \( X \sim N(0, 1) \), limit \( X \)nononoyes
independent, \( P(X_n = n) = \frac{1}{n} \)no (BC II)yesno (\( L^r \) iff \( r < 1 \))yes
  • Borel–Cantelli: \( \sum P(A_n) < \infty \Rightarrow P(\text{i.o.}) = 0 \); independent and \( \sum = \infty \Rightarrow P(\text{i.o.}) = 1 \).
  • Slutsky: \( X_n \xrightarrow{d} X \), \( Y_n \xrightarrow{P} c \) give \( X_n + Y_n \to X + c \), \( X_nY_n \to cX \), \( X_n/Y_n \to X/c \) (\( c \ne 0 \)).
  • WLLN: Chebyshev (uncorrelated, bounded variances); Khintchine (i.i.d., finite mean).
  • SLLN (i.i.d.): \( \bar{X}_n \to \mu \) a.s. iff \( E\lvert X_1 \rvert < \infty \). Apply to \( g(X_i) \): \( (X_1\cdots X_n)^{1/n} \to e^{E\log X} \) (\( \frac{1}{e} \) for uniform).
  • CLT (i.i.d., \( 0 < \sigma^2 < \infty \)): \( \frac{S_n - n\mu}{\sigma\sqrt{n}} \xrightarrow{d} N(0, 1) \); not in probability. Continuity correction \( \pm\frac{1}{2} \).
  • Delta method: \( \sqrt{n}(g(\bar{X}_n) - g(\mu)) \xrightarrow{d} N(0, g'(\mu)^2\sigma^2) \).
  • Cauchy: \( \bar{X}_n \) is standard Cauchy for every \( n \).
\( z \)11.51.6451.9622.53
\( \Phi(z) \)0.84130.93320.95000.97500.97720.99380.9987

Last-week checklist

Back to the chapter overview, or revise the analysis this chapter rests on with Real Analysis: Quick Revision.

Advertisement
Advertisement