ED+A year 2

Question 1

Null hypothesis

Answer 1

no significant difference between specified populations. Hypothesis about a POPULATION.

Question 2

Statistical inference

Answer 2

assessing what we can say about the population on the basis of our knowledge of the sample - sampling with replacement is the usual situation

Question 3

Sampling with replacement

Answer 3

Population is so large relative to the sample that the probability of the next individual sampled being of a particular type is independent of the make-up of the sample so far

Question 4

Reject null hypothesis if…

Answer 4

Probability of the sample deviating as much as it does (or more) from the null hypothesis given that the null hypothesis is true for the population is SMALL (large deviation is only expected in less than 5% level of significance symbolised by alpha)

Question 5

Random sampling

Answer 5

the probability of inclusion of an individual from the population in the sample is the same for all individuals, and independent of the rest of the sample

Question 6

2 forms of non random sampling

Answer 6

Bias and Non-independence

Question 7

Bias non-random sampling

Answer 7

probability that an individual from the population being included in the sample depends on the individuals characteristics with respect to the trait being studied e.g. males are more likely to go out than females and be more likely to be spotted therefore are more likely to be included in the sample.

Question 8

Non-independence non random sampling

Answer 8

probability of an individual from the population being included int he sample is independent of its characteristics, but is dependent on other members of the sample. e.g. if animals move in single sex flocks sampling an individual male might always be associated with sampling a further half dozen

Question 9

Type I error and when it is most concerning

Answer 9

Rejection of correct null hypothesis in 5% proportion of tests. Most concerning when doing multiple simultaneous testing e.g. testing for the frequency of many different diseases in people living near transmitters and people not - if in reality there was no disease caused we would expect for five diseases frequencies would be higher in those living near and not living near.

Question 10

Preventing type I errors and the consequences

Answer 10

Bonferroni correction - if we needed a probability (alpha) of 5% in order to reject a null and we did 200 tests, the bonferroni correction would require us to look for a probability of 5%/200 (0.025%) in order to reject any null. Consequence - if a null is true, we have only a 5% chance of getting a single significant result in all the tests that we do