critical appraisal & hypothesis testing

Question 1

parametric tests

Answer 1

certain rules (assumptions) to follow (e.g. sample size, normality, linearity), if these assumptions are not met, alternatives should be used

Question 2

non-parametric tests

Answer 2

alternative significance tests which do not require assumptions to be met in order to be performed (any sample size, distribution etc.)
can be used at any time but not as powerful

Question 3

null hypothesis

Answer 3

relates to no difference , aim to have enough evidence to disprove this

Question 4

probability

Answer 4

range from 0-1
0= definitely not going to happen
1=definitely going to happen

Question 5

p-value

Answer 5

a probability applied to significance testing

Question 6

region of interest

Answer 6

• The probability of a specific value is very small and not very useful
• Instead we talk about a range of value (our observed value or more extreme)
• If our study estimate was more extreme (even bigger difference), I would not change my mind (my overall conclusion would not change
• If I observe a value that is close to my null hypothesis, my P-value will be large (close to 1)
• Therefore it s likely that any difference observed is just due to chance (there is no significant difference between observed and null value
• So the null hypothesis appears to be true

Question 7

p =0.05

Answer 7

often used as a cut off for statistical significance

Question 8

relationship between confidence intervals and p-values

Answer 8

if the null is not in the 95% CI, your p-value will be less than 0.05