Probability space — sample space Ω, sigma-algebra F, measure P — Kolmogorov axioms: Visual walkthrough

4.9.1 · D2 · Maths › Probability Theory & Statistics › Probability space — sample space Ω, sigma-algebra F, measure

Kisi bhi symbol se pehle, ek vaada: yahan sab kuch sirf area hai. Poore sample space $Ω$ ko (yani sab kuch jo ho sakta hai) ek aisa rectangle samjho jiska total area exactly 1 unit hai — ek liter paint. Ek event us rectangle ke andar ek region hai. Uski probability us par padi paint ka fraction hai. Yeh ek mental image poore page ko sambhale rakhti hai.

Step 1 — Stage: area 1 ka ek rectangle

KYA. $Ω$ ko ek rectangle ki tarah draw karo. Uska area $1$ rakho. Koi bhi event $A$ uske andar ek blob hai.

KYUN. Teen axioms sirf numbers $P (A)$ ki baat karte hain. Unhe dekhne ke liye humein ek aisi picture chahiye jahan woh numbers visible quantities hon. Area perfect hai: yeh kabhi negative nahi hota, jab regions overlap na karein toh add hota hai, aur ek fixed total ko $1$ tak normalize kiya ja sakta hai. Yahi exactly axioms A1, A3, A2 hain doosre roop mein — isliye "area" koi dheela analogy nahi hai, yeh ek faithful model hai.

PICTURE. Figure dekho. Poora peach rectangle $Ω$ hai; magenta blob $A$ hai.

Figure — Probability space — sample space Ω, sigma-algebra F, measure P — Kolmogorov axioms

Step 2 — Adding rule: disjoint pieces add ho jaate hain

KYA. Agar do events $A$ aur $B$ kabhi overlap nahi karte (koi bhi outcome share nahi karte), toh " $A$ ya $B$ " ka area sirf dono areas ka sum hai.

KYUN. Yeh axiom A3 hai, countable additivity, sabse simple case mein draw kiya gaya. "Disjoint" ka matlab hai ki blobs kahin bhi touch nahi karte; tab unhe paint se cover karne mein ek blob jitni plus doosre jitni paint lagti hai — koi paint do baar count nahi hoti kyunki koi region dono mein nahi aata.

PICTURE. Do alag alag blobs; combined shaded area literally sum hai.

Poora axiom A3 infinitely many disjoint pieces $A_{1}, A_{2}, \dots$ ko add karne ki ijazat deta hai. Humein sirf do wali picture chahiye, lekin peeche dimaag mein rakhna: yeh rule tab bhi kaam karta hai jab tum $Ω$ ko countably infinite non-overlapping crumbs mein kaato.

Step 3 — Empty event mein koi paint nahi hoti

KYA. Dikhao ki $P (\emptyset) = 0$ : woh region jisme koi outcome nahi hai uska area zero hai.

KYUN. Yeh obvious lagta hai, lekin koi bhi cheez obvious nahi hoti jab tak axioms se force na ho. Trick yeh hai: empty region $\emptyset$ apne aap se aur kisi bhi cheez se disjoint hai. Toh main $\emptyset$ ki infinitely many copies $Ω$ ke saath stack kar sakta hoon aur union phir bhi sirf $Ω$ hi rahega.

PICTURE. Rectangle $Ω$ , phir infinitely many zero-width slivers ki ek stack (har ek $\emptyset$ ki copy) jo total mein kuch nahi jodtey.

Step 4 — Poore ko split karna: complement rule

KYA. $Ω$ ko ek blob $A$ mein aur uske bahar ki har cheez mein split karo, jise $A^{c}$ kehte hain ("A-complement", bacha hua region). Unke areas ka total $1$ hona chahiye.

KYUN. $A$ aur $A^{c}$ disjoint hain (koi cheez andar aur bahar dono mein nahi ho sakti) aur saath milkar poore rectangle ko tile karte hain. Toh Step 2 ka adding rule apply hota hai, aur total A2 se $1$ par fix hai. Yeh probability ki sabse zyada use hone wali rule hai: "A nahi hone ki chance."

PICTURE. Rectangle magenta $A$ aur violet $A^{c}$ mein cut hua; dono colors ise exactly bharte hain.

Step 5 — Nesting: bada blob, zyada paint (monotonicity)

KYA. Agar $A$ poori tarah $B$ ke andar hai (likha jaata hai $A \subseteq B$ ), toh $A$ ka area $B$ se zyada nahi ho sakta.

KYUN. $B$ ko outer blob aur $A$ ko inner blob draw karo jo usmein poori tarah contained hai. Unke beech ki ring, $B ∖ A$ (" $B$ minus $A$ ", $B$ ka woh hissa jo $A$ ke bahar hai), ek asli region hai jiska apna non-negative area hai. Kyunki $B$ , $A$ aur us ring ka combination hai, aur ring area subtract nahi kar sakti, isliye $B$ kam se kam $A$ jitna bada hai.

PICTURE. Nested blobs: inner $A$ (magenta), outer $B$ , orange ring $B ∖ A$ unke beech mein.

Step 6 — Overlap: subtract kyun karna padta hai (inclusion–exclusion)

KYA. Jab $A$ aur $B$ overlap karte hain, unke areas ko add karne par overlap $A \cap B$ double-count ho jaata hai. Ise theek karo us overlap ko ek baar subtract karke.

KYUN. Yeh woh case hai jo Step 2 ne mana kiya tha. $A$ par paint daalo, phir $B$ par: lens-shaped region jahan woh cross karte hain ( $A \cap B$ , " $A$ aur $B$ ") do baar paint ho jaata hai. Union par paint ek baar count honi chahiye, toh double-counted lens ki ek copy hatao.

PICTURE. Do overlapping blobs; beech ka lens double count dikhane ke liye darker shaded hai jise subtract karna hai.

Step 7 — Degenerate case: zero area ka ek point jo phir bhi occur ho sakta hai

KYA. Ek continuous stage jaise $Ω = [0, 1]$ par, ek single point ka area $0$ hota hai, phir bhi woh occur ho sakta hai. "Probability $0$ " ka matlab "impossible" nahi hai.

KYUN. Point $0.5$ ke around ek tiny interval ko shrink karo. Har interval $[0.5, 0.5 + \frac{1}{n}]$ ka area (length) $\frac{1}{n}$ hai, aur $\frac{1}{n} \to 0$ jab $n$ badhta hai. Point in shrinking intervals ka limit hai, isliye uska area limit $0$ hai. Lekin jab tum line par dart phenkte ho toh tum exactly $0.5$ par land kar sakte ho — outcome real hai, uska area nil hai.

PICTURE. Number line $[0, 1]$ jisme nested intervals point $0.5$ par collapse ho rahi hain, unki lengths $\frac{1}{1}, \frac{1}{2}, \frac{1}{3}, \dots$ $0$ ki taraf ja rahi hain.

Ek-picture summary

Is page par har rule total area 1 wale rectangle ke andar area ke baare mein ek statement hai:

disjoint pieces add karo (Step 2),
bacha hua poora complete karta hai (Step 4, complement),
nested matlab bada nahi (Step 5, monotonic),
overlap double-count hota hai, toh subtract karo (Step 6),
aur kuch nahi tak shrink karne par area $0$ milta hai (Steps 3 & 7).

Recall Poore walkthrough ki Feynman-style retelling

Ek rectangular table top imagine karo jispar exactly ek liter spilled paint hai — woh hai $Ω$ total probability $1$ ke saath. Table par jo bhi shape draw karo woh ek event hai, aur uski probability bas itni hai ki kitni paint uske andar baithi hai. Rule ek: negative paint nahi ho sakti. Rule do: poori table par sab ek liter hai. Rule teen: do shapes jo kabhi touch nahi karte — unki paint bas add ho jaati hai. Sirf is se: ek empty shape mein koi paint nahi; ek shape ke bahar ki paint ek liter minus andar ki paint hai; ek bade shape ke andar ka shape bade wale se zyada paint nahi rakh sakta; aur agar do shapes overlap karein to overlap ko do baar paint kiya, isliye ek baar subtract karo. Aakhir mein, ek smooth table par ek single dot ka koi area hi nahi, isliye koi paint nahi — lekin tumhari ungli phir bhi us par land kar sakti hai. Yahi hai basic probability ka poora khel, aur yeh sab sirf area hai.

Recall Rapid self-test

$P (\emptyset) = 0$ kyun force hota hai? ::: Infinitely many empties ko $Ω$ ke saath stack karo; union phir bhi $Ω$ hai, isliye $P (\emptyset)$ terms ka pile $0$ mein sum hota hai, aur har non-negative term $0$ honi chahiye. $P (A) \leq 1$ kahan se aata hai? ::: Complement rule $P (A) + P (A^{c}) = 1$ se, jisme $P (A^{c}) \geq 0$ hai. $P (A \cup B) = P (A) + P (B)$ exactly kab hold karta hai? ::: Sirf jab $A \cap B = \emptyset$ ho; warna $P (A \cap B)$ subtract karo. Kya $P ({0.5}) = 0$ ka matlab hai ki $0.5$ , $[0, 1]$ par occur nahi ho sakta? ::: Nahi — probability $0$ ka matlab hai "almost never", point phir bhi ek possible outcome hai.

Connections: parent 4.9.01 Probability space — sample space Ω, sigma-algebra F, measure P — Kolmogorov axioms (Hinglish) · builds toward Conditional probability & independence · Random variables · Inclusion–exclusion principle · Borel sigma-algebra · Measure theory & Lebesgue integration · Expectation and Lebesgue integral.