Flashcards
Lecture 7
Cards in this deck (14)
- 1What is probability sampling and what type of sample is usedA simple random sample is used as this will mean that each sample will have the same probability of being selected and the sample should represent the entire population
- 2What is stratified sampling?This is sampling used when the population comprises of different types of similar individuals.
- 3What is a stratum?This is a population sub division of similar units
- 4What do you do in stratified sampling which creates more of a representative of a sample?You take a simple random sample from each stratum. And sometimes since proportions are heard to match, we can take different sized samples from different strata do reduce the overall variability
- 5What is cluster sampling?This is where a you take a simple random sample of groups.
(single stage)=sample all the units in the selected groups
(two stage)=take a simple random sample of units within a group - 6What is replication and randomisation?This allows us to separate out treatment effects from chance effects
- 7What is randomisation?This ensures that effects of unmeasured factors are equalised across the treatment groups. Ie differences equalised.
- 8What is a completely randomised design?Where there is no restrictions on the allocations apart from the numbers in each group
Each possible allocation has the same probability of being selected - 9What is a randomised block design?This is the random allocation of treatments within a block of similar subjects. It reduces unexplained error by moving.
(block what you can, randomise what you cannot) - 10What are repeated measures?This is where an individual can act on their own block. Eg sunscreen testing where both gets tested on that one individual
- 11What is a paired T test?This is a special case of analysis where just 2 treatments are used.
- 12When does confounding occur?Confounding occurs when the analysis of the data gives an overestimate or underestimate of the true value because other variables are not correctly accounted for
- 13How do we deal with confounders?We do this through randomising in experiments to randomise/equalise all other effects across the treatment groups.
- 14What may go wrong in a good study?Where if there is a survey non response (low response rate from surveys)
There has been a recording error, treatments failed, study plots destroyed
Information
- Description
- University Statistics Flashcards on Lecture 7 , created by arielhchen on 02/08/2015.
- Created by
- arielhchen
- Views
- 0
- Rating