TOPIC #8: Sample surveys

Module 3, DATA1001

how a population is surveyed

census

sample is a…

subset of a population

limitations of a census

• time consuming to collect information about every unit of a population

• resource intensive

• costly

• difficult to enforce participation.

parameter

numerical fact about the population which we are interested in e.g population mean

estimate (or statistic)

calculation of sample values which best predicts the parameter, e.g sample mean

selection bias

systematic tendency to exclude or include on type of person from the sample

non-response bias

caused by participants who fail to complete surveys

interviewer bias

when the interviewer has to make a choice of particpants in the survey or when characteristics of the interviewer have an effect on the answer given by participants

measurement bias

when the form of the question in the survey affects the response to the question.

examples of measurement bias

recall bias: people forget details

social desirability bias: people may not tell the truth

lack of clarity in the question: people may misinterpret the question, or certain words may be ambiguous and so mean different things to different people.

simple random sampling

probability sampling technique which involves drawing individuals from a population at random without replacement - each individual has an equally random chance of being selected from the population

stratified sampling

sampling from a population that is partitioned into subpopulations (e.g gender, age, location, etc). sampling is done independently in each subpopulation (stratum), ensuring that the sample is representative of the population with respect to certain characteristics.

multistage cluster sampling

takes samples in stages and simple random sampling is performed within each stage.

quota sampling

non-probabilistic version of stratified sampling, assembled sample is required to have a pre-determined number of individuals with respect to certain characteristics

may result in unintentional bias from the interviewers when they choose subjects to survey

convenience sampling

non probability sampling technique, subjects are selected because it is convenient to select them. (not recommended as the partcipants that are easy to recruit may not be representative of the population of interest)

sample surveys always involve ….. error, because…..

chance, each sample is just one possible draw from the population.

why does the SE from the box model need to be adjusted to get the exact SE for a survey?

because in practice surveys are drawn without replacement and the same person can’t be interviewed twice.

correction factor (finite population correction)

SE_{without replacement}= correction factor x SE_{with replacement}

where:
correction factor = root [(population size - sample size) / (population size - 1)]

how do you construct a confidence interval from a single sample survey?;

1. estimate the box population proportion (mean) to be equal to the sample proportion.

2. follow the box model and the CLT, we can then assume the sample proportion (mean) follows a normal distribution, eith EV = mean_box, and SE = SD_box/root n)

bootstrapping

a computational process that allows us to estimate the properties of the population, by using the properties of a particular sample. it is a method of resampling with replacement from a sample to estimate variability and CIs.

what is a confidence interval

A range of plausible values for the population parameter based on sample data