Flashcards in Lecture 5 - Sampling Procedures Deck (42):
Define target population
It is the totality of the cases that conform to some designated specifications such as demographics or psychographics, or say the population of a country
What is a census?
Surveying the entire target population, which is costly, time consuming, and sometimes impossible
Instead of implementing a census, it's best to what?
take a sample from the target population
What is the 6 step sampling procedure?
1. Define target population
2. identify the sampling frame
3. Select a sampling procedure
4. Determine sample size
5. Select sample elements
6. Collect data from selected elements
What is a sampling frame?
It is the list of people/units who are eligible in the target, and the sample is drawn from them
What are the 3 types of sampling?
1. Random Sample
2. Stratified sample
3. Cluster sample
What is unit of analysis?
It is what you are studying, so could be students, store, households etc.
(This is when you don't know probability of being included in sample)
(Each population element has a known chance of being included in the sample)
With sampling, what is incidence rate?
# target pop / # whole pop
So if you are only targeting small amount of pop the incidence rate will be tiny
What are 3 types of non-probability samples?
1. Convenience Samples
(elements are chosen because they're convenient and available easily)
2. Judgement Samples
(elements handpicked by researcher)
3. Quota Samples
(selecting sample that looks like population by using some criteria)
What are judgement samples suitable for and method?
Suitable for exploratory research, subjects that can be experts selected.
Snowballing is a method, which is a special group of subjects identified, then these people help you find more people like them
What kind of sample are most of the projects in this class going to use?
Convenience samples, because a lot of people are going to be putting their surveys on Facebook and whatnot because we don't have money/resources for random sampling
What is an example criteria for quota samples?
So you could select sample by using an observable criteria such as gender, or using unobservable such as personality
Or if you wanted something like
Than you could do survey with screening question to make sure you fulfill quota of percentages. So once you get 30 marketers the quota is filled and no more from marketing is needed/wanted
If you were randomly talking to people in the Rowe using quota sampling, do you need a sampling frame?
Nope, you don't need a sampling frame because you don't have the list before hand or anything, you're just going around asking people stuff
What is a bad part of quota samples?
Well you might choose quota on some criteria, but you might miss criterias. So choosing to select based on Uni majors criteria, but ignore intl students criteria.
Explain what happens with a simple random sample?
Every population element is equally likely to be selected. So every sample with size n is equally likely to occur as any other sample with size n
What is a super simple way to draw a random sample?
1. Give id numbers to each population element
2. Find id numbers from a list of random numbers
What is a parameter?
It is a characteristic or measure of a population.
So things like Canada pop average income level. So it's a population parameter that could be compared to say, the US.
N: size of population
u: population mean
o (sigma): population standard deviation
n= sample size
x (with bar onto): sample mean
s = sample standard deviation
Define sampling distribution?
So this is all the possible samples of size n that can be selected from a population of size N
What is the central limit theorem?
So when n is large, the sampling distribution will be NORMAL with mean equal to u (pop mean) and standard deviation equal to (formula)
So it shows a nice normally distribute curve for a sampling distribution, that the farther away from the pop mean, the less frequency of samples.
What are the three z for confidence interval estimates?
z = 1, 68.3% of all sample means are within range
z = 2, 95.5% of all sample means are within range
z = 3, 99.7% of all sample means are within range
So z = 1 would be most precise out of these, but you would also only be confident 68.3% of the time
If the standard deviation of the population is unknown, what are the two things you can do?
1. Use standard dev or population from other research projects that measured the same variable
2. Assume standard deviation of population equals sample standard deviation
What is stratified sampling?
Random sample is selected from each stratum
What is a stratum (plural is strata)
it is mutually exclusive and collectively exhaustive subsets of a population
Give an example of a mutually exclusive and exhaustive stratified sampling
so for age
For stratified samples, how are they weighted? (2 of them)
proportionate stratified sampling
(So each stratum size is proportionate to size in population)
Disproportionate stratified sampling
(Size of sample from stratum is determined also according to the variation within that stratum)
How is disproportionate stratified sampling done with regards to variation within stratums?
So the greater variation within a stratum, the more sampling from it.
What are 3 reasons to use stratified sampling?
1. Lower variability within strata increases precision of the sample mean
2. Investigate some characteristics of particular subgroups
3. Ensure representativeness of important segments
What are the two steps wit cluster samples?
1. divide target population into clusters that are mutually exclusive and collectively exhaustive
2. Randomly select a subset of clusters
What must clusters be?
Representative of target population
What are one and two stage cluster sampling?
One-stage cluster sampling:
use all elements in the selected cluster
two-stage cluster sampling:
randomly draw a sample from each cluster
Give example of one-stage cluster sampling with wine consumption in Chicago
So we randomly select say 100 blocks out of 10,000 blocks in Chicago.
Then we measure wine consumption fro all households in the selected 100 blocks
What is a pro and con with one-stage cluster sampling?
1. Less costly, because easier after clustering
2. Less representative, because households in say a cluster could be homogenous, and thus subsets are not representative
What are the two stages of two-stage cluster sampling?
1. Randomly select n clusters out of N population clusters available
2. Randomly select sample k from each cluster
With two-stage cluster sampling:
n*k = desired houlseholds to survey with Chicago wine example
Greater n and smaller k means:
Smaller n and greater k means:
Greater n and smaller k means:
(more costly and more representative)
Smaller n and greater k means:
(less costly, and less representative)
For the objective factor, what is the difference between stratified and cluster sampling?
For the subpopulations factor, what is the difference between stratified and cluster sampling?
all strata included
sample of clusters chosen
For the within subpopulations factor, what is the difference between stratified and cluster sampling?
each stratum should be homogeneous
each cluster should be heterogeneous
For the across populations factor, what is the difference between stratified and cluster sampling?
strata should be heterogenous
cluster should be homogeneous
For the sampling frame factor, what is the difference between stratified and cluster sampling?
needed for entire population
needed only for the selected clusters