sampling

sampling is about

• understanding the population characteristic
• controlling randomness,

Some of other key points are bias, cost, time and representativeness

handling cost is about

• repeating experiment (which is costly) vs bootstrapping (resampling the samples, and get a distribution, then calculate some stats)
• selecting a sampling technique

dividing to N-1 instead N?

• sample ν is always smaller than the population ν. So to compensate the gap, divide N-1

 sampling in two main areas  
 
 population    | data gathering
 model         | train-test split

sampling techniques

   random      |  
   convenient  | easy picks    
   clustering  | then, random picks | grouping and selecting only from some clusters !
   strata      | then, random picks | grouping and selecting from all stratum !

note

• cross validation is a kind of bootstrapping
• ensemble a model is also a kind of cross validation