Sampling Flashcards
Name 3 types of non-probability sampling
- Judgement
- convenience
- purposive → subjects chosen blc they have certain attribute
probability sample
each person has a non-zero probability Of being included in study
simple Random Sampling
→ sample fixed proportion
→ can be difficult to implement
→ same probability of being picked
→ Random # generator
Systematic Random Sampling
1.determine sample size
2. Determine sampling interval by dividing pop Size by sample size
3. Randomly select start point from within first interval
4. sample each person after based on sampling internal
ex. in a pop of 1000 chose every 10th person can also be time- every 12 hours.
→Simple, inexpensive
All units can be counted , calculate internal based on sample size t poprandomly select 1st
Stratified Random sampling
1 . total pop is divided Into subgroups Then ppl are randomly selected in each stra
Cluster sampling
collection of groups with one or more characteristics in common
- randomly select# of clusters(primary sampling unit)
- every person in clusters is sampled( unit of concern)
→ variance btwn groups greater so higher Standard error. ‘ !
Multi stage s ampling
similar to cluster sampling G useful if cluster has too many ppl to sample or if ppl in Cluster are too alike
- Randomly select PSU
- selected secondary sample
→ to ensure all individuals in pop have some probability select you can do 1 of 2 things:
a) Select PSU with probability proportional to their size
ex: if class size is known, larger classes will have a higher probability of being selected → select PSU’s then select fixed# of students to be sampled in each class
b) if class size is not known, then select a simple random sample of PSU’s then sample a constant proportion of students in each class .
what is the multistage sample size calculation based on?
the relative cost of sampling and the between-class variation if btwn-class variation is high. more classes will need to be sampled.
type 1 error
outcomes in groups are declared different when in fact they are not
Type 2 error
atcomes in groups being compared are declared not different when in fact they are
power
probability of finding a stat sig difference when it exists and is of somemagnitude
confidence
degree of certainty that what we have observed is real and not due to chance
Allowable error
describes acceptable analytic performance
Target population
pop where we extrapolate results to
ex All households in a region
source pop
pop where study units are drawn from -→ there should be a conceptual list of all members and non-zero probability of being chosen
ex: households eligible /participating