Friday, March 4, 2011

Sampling: Design and Analysis – Sharon Lohr – Chapter 4 Stratified Sampling

 

What is stratified sampling

 

-          Supplementary info to help design sample

o   Men generally earn more than women

-          If variable we are interested in takes on different mean values in different sub populations

-          Reasons

o   Protection from possibility of obtaining a really bad sample (randomly chosen but not representative)

o   Want data of known precision for subgroups

o   Cost and convenience considerations

o   Stratified sampling, if done correctly, will give more precise estimate for whole population

 

Allocating Observations to Strata

 

-          Proportional allocation

-          Optimal allocation

 

 

Post Stratification

 

-          Avoid data snooping

 

Quota Sampling

 

-          Many samples that masquerade as stratified random samples are actually quota samples

-          Population divided into different sub populations

-          Important difference à probability sampling is not used to chose individuals from sub populations for sample

-          Don't know what is selection bias

 

 

 

No comments: