Sampling Flashcards
population
everyone fitting criteria
clearly defined
sample
subset of pop
representative of whole pop
minimise sampling error and variability
characteristics of ideal sample
representative
independence
adequacy
homogeneity
2 categories of sampling methods
probability sampling (random)
non probability sampling
probability sampling
simple random sampling
stratifed
systematic
multistage
non probability sampling
judgment or deliberate
convenience
snowball
quota
probability sampling advantages
requires detailed info to be effective
estimates which can be measured precisely and unbiased
can evaluate relative effectiveness
probability sampling disadvantages
high degree of skill
requires time to plan
costs are higher
simple random sample
random number generated
random sample +ve and -ve
+
simple
no bias
representative
easy to detect errors
-
lack of control
stratified
ratio of number of items to be selected from each strata should be the same as the total number of the units in the strata bearing the units of entire pop
types of stratified
proportionate
disproportionate
stratified weighted
stratified + and -
+
greater control
representative
can replace units
-
bias
hard to achieve proportion
hard to make representative
hard to place cases under level
stratified + and -
+
greater control
representative
can replace units
-
bias
hard to achieve proportion
hard to make representative
hard to place cases under level
systematic + and -
+
Simple to draw
variance are smaller
reduces variability
-
interval related to ordering may increase variability
estimates of error likely
multistage sample
random sampling in each sampling stages where there are more than 2
multistage sample + and -
+
complete list of pop not needed
lower cost of travel f geographically defined
-
errors likely to be large in simple or systematic
errors increase as no. of selected sample units decrease
non probability sampling methods
do not provide every item in pop with equal chance
partially subjective
judgment sampling
choice of sample depends primarily on judgment of researcher
judgment sampling + and -
+
enables inclusion of important units
obtains more representative when looking at unknown traits
practical
-
risk of including researchers needs
can objectively evaluate reliability
convenience sample
when population not well defined
sampling unit not clear
complete list not available
snowball sample
research contacts with small no. of ppl who establish new connections
snowball sample problems
no accessible sample frame
unrepresentative
quota sample
- classify pop into types that assumed to be relevant
- determine proportion of population falling into each type
- setting of quotas for each interviewer who has responsibility of selecting ppl so contains correct proportion
quota sample + and -
+
low costs to prepare
some stratification effect
-
bias of investigator
errors cannot be estimated by stat procedures
factors affecting reliability
size
representativeness
parallel sampling (another sample for testing)
homogeneity of samples (have all same features)
unbiased selection
3 types of errors
sampling variability
sampling error
non sampling error
sampling variability
different samples from the same pop don’t always produce same mean and sd
sampling error
mean of sample won’t be same as mean of pop
non sampling error
eg
questions asked badly
poor measurements
errors in recording