| Question | Answer |
| What're the sampling methods? | -random sampling
- non random sampling
- probability sampling
- non probability sampling |
| What is inference? | the act of passing from statistical sampling data to generalization, usually with calculated degrees of certainty |
| Why is using a sample very common? | small portion of population costs less than the entire population, needed if the population is large, little time/budget. Sampling is destructive |
| Probability sampling | - each element has equal probability of being selected, needed if you want a representative sample
- not easy, need a good sampling frame |
| non-probability sampling | - relies on judgement
- probability to be selected unknown |
| Probability sampling methods | 1. simple random sampling (computer select from frame)
2. Systematic sampling
3. Stratified sampling, select strata of interest + sample from them w method 1/2
4. cluster sampling, randomly select area + get info from sample , reduces travel costs |
| non- probability sampling methods | 1. Convenience sampling: Select people who are available
2. Judgmental sampling: select those who are judged
3. Quota sampling: match characteristics of sample to population
4. snowball sampling: let respondents recruit others |
| What're resulting errors when working with a sample? | - sampling error: difference between true and sampling information (not representative for population)
-non sampling error: measurement error, non response |
| How bad is non-response? | - Coincidental nonresponse: respondents + nonrespondents dont systematically differ on imp variables. may give right image, need a bigger sample
- systematic nonresponse: differ on imp variables, gives wrong image, need bigger + better sample |
| Check representativeness with smart comparisons | 1. compare respondents to population data, secondary data
2. compare respondents to non-respondents with central question |
| How to cure non-response | 1. Methods before data collection: prior notification, give incentives
2. Methods during data collection: Re-invite, approach again
3. Methods after data collection: incl socio-demographics in analysis, weighing (underrepresented categories more weight) |
| How to make sample more representative | Weighing on important variables |
| What factors does sample size depend on? | more is better. why?
- required precision
- confidence in statements
- heterogeneity of the population
- size of the population
- need for segments
- time, budget etc |
| Confidence intervals | When estimating a mean, theres always uncertainty, confidence interval shows how much uncertainty. When the data collection is repeated 95% of sample holds true value.
More observations means narrower confidence interval around the mean |
| Limitations of online surveys | - no internet access
- phone screen small for some question types
- online invite easy to dismiss
- clicking answers is easy so maybe to quick responses
-control question may show many people not paying attention |
