Topic 8: Sample Surveys

Question 1

Differentiate parameters and estimates (statistics)

Answer 1

• Parameters: numerical facts about the population which we want to know
• Estimates (statistics): calculation of the sample values which is known and used to best predict the parameters

Question 2

Name of sampling method

Answer 2

• Multi-stage cluster sampling: take samples in stages and choose individuals/clusters randomly at each stage (probability sampling)
• Quota sampling (non-probability): samples has the same proportion of individuals as the entire population with respect to known characteristics, traits, phenomenon => unintentional bias
• Convenience sampling (non-probability): select subjects based on their convenient accessibility
• Sample random sampling (probability): interviewer not involved and can compute the chance of any individual being chosen

Question 3

What does estimate equal to?

Answer 3

Estimate = Parameter + Bias + Chance error or
Estimate = Parameter + Non-sampling error + Sampling error

Question 4

What is box model used to quantify?

Answer 4

Box model is used to quantify the likely size of CE os estimated proportion from sample modeling.
SE measures variability across different samples.

Question 5

How does sampling with and without replacement affect accuracy of SE?

Answer 5

• With replacement: SE relies on absolute sample size
• Without replacement: increasing ratio of sample to population size will decrease SE (increasing proportion of population being sampled means the variability decreasing)

Question 6

Why does SE in sample surveys needs to multiply by the correction factor?

Answer 6

Sample surveys are sampling without replacement, so the SE cannot be determined solely on the size of the sample.

SE (w/out) = sqrt((N-n)/(N-1)) * SE(w/t)

Question 7

Explain bootstrapping

Answer 7

Bootstrapping is to estimate properties of the population by using those of the known sample

Question 8

What is confidence interval?

Answer 8

Each sampling will give a different value to the estimate we are looking for, so we need a range in which the true value of the population will lie with a high probability.

68% CI = sample proportion +/- 1SE
95% CI = sample proportion +/- 2SE
99.8% CI = sample proportion +/- 3SE