BASE RATE
Shabdon ka matlab, aur har shabd ki limit.
Is section ke har shabd ke 2 hisse hain: uska matlab, aur woh kya nahi batata. Doosra hissa hi asli kaam karta hai, kyunki sirf matlab padhne wala reader us shabd par bharosa karna seekh jaata hai. Yahan wahi shabd hain jinme har study likhi gayi hai, un limits ke saath jo unke peeche chalti hain.
Har entry ke 2 hisse hain: matlab, aur woh kya nahi batata.
Base rate khud ek anumaan hai, uska apna sample aur apni avadhi hoti hai.
Effective sample raw ginti se chhota hota hai, aur wahi asli ginti hai.
Interval number se zyada maayne rakhta hai, khaas kar jab woh shoonya ko chhoo raha ho.
BUNIYAD
Base rate aur uska control
Base rate yeh hai ki koi cheez pehle kitni baar hui — bahut saare purane cases par gina gaya, aur uske saath aane wali uncertainty ke saath likha gaya. Woh agle mangalwar ke baare mein kuch nahi kehta, kyunki agla mangalwar us sample mein tha hi nahi.
Matched control usi date ke woh securities hain jo khud shart poori nahi karte the, lekin price band, volatility band aur traded-value band mein us event jaise hi the. Yeh control usi cheez par kaam karta hai jis par match kiya gaya ho — sector in keys mein nahi hai, isliye ek poore sector ke ek saath hilne ko yeh pakad nahi paata.
GINTI
Sample, design effect aur interval
Effective sample size woh ginti hai jitne sach much azaad observations sample mein hain. Ek hi din ke events ek hi bazaar ki chaal baant lete hain, isliye raw ginti sample ko bada dikhati hai. Design effect wahi paimana hai jo batata hai ki ek din ka clustering sample ko kitna chhota kar deta hai.
Confidence interval woh daayra hai jiske saath measurement mel khata hai. Woh sirf sampling ki uncertainty batata hai, yeh nahi ki duniya khud badal gayi ya nahi. Interval agar shoonya ko apne andar rakhta hai, toh is sample ne pattern ko bazaar se alag kar ke dikhaya hi nahi.
UNIVERSE
Kaunse naam sample mein aate hain
Point-in-time universe ka matlab hai ki kisi din ki eligibility usi din ke published record se tay hoti hai. Jo securities baad mein delist, merge ya suspend ho gayin, woh un saare dinon ke liye sample mein rehti hain jab woh trade ho rahi thi. Sirf aaj listed naamon par study karna wahi galti hai jo apne aap ek jhootha edge bana deti hai.
Censoring un observations ke liye hai jinki security window khatam hone se pehle band ho gayi. Unhe hataya nahi jaata. Unhe aakhri traded price par band kar diya jaata hai aur alag se nishaan laga diya jaata hai.
SEEMA
Kuch shabd jo aksar zyada bharosa de dete hain
- Pre-registration natije ko sahi nahi banati. Woh sirf natija dekh kar setting chunne se rokti hai.
- Knife edge woh natija hai jo ek setting par dikhta hai aur uske bagal wali settings par gayab ho jaata hai.
- Placebo check sirf yeh batata hai ki method khud se natija nahi bana raha.
- Insufficient sample ka matlab pattern ka na hona nahi hai — matlab yeh sample bata hi nahi sakta.
Kitne observations kaafi hote hain
In shabdon mein sabse zyada gumraah karne wala sample size hai — bada percentage chhote sample par bhi bada dikhta hai, aur wahi page yeh farak khol kar rakhta hai.