REBM Flashcards
What is the null hypothesis?
Predicts no effect or relationship between variables (Ho)
What is the alternative hypothesis?
Predicts there is an effect or relationship between variables (Ha)
Sometimes called the study hypothesis
What are one tailed and two tailed tests?
One tailed= testing for the possibility of the relationship in one direction, disregarding the possibility of another direction
Two tailed= testing for the possibility of a relationship in either direction
Is it easier to achieve statistical significance in a one or two tailed test?
One- the p value is not split in either direction
What is p hacking?
Where lots of tests are ran on two sets of data with the view of getting a statistically significant value by luck. Considered highly unethical
What is the p value?
A statistical measurement used to determine the likelihood that an observed outcome is the product of chance
‘The probability of obtaining results as least as extreme assuming the null hypothesis is correct’
What does a p value of 0.05 mean?
5% of the time would see a result as least as extreme as the one you got if the null hypothesis was true
Can the p value ever be 0?
No- always a chance results were a coincidence
What is a type 1 error?
Wrongly rejecting the null hypothesis when it is true
What is a type 2 error?
Wrongly failing to reject the null hypothesis when it is false
Can you accept the null hypothesis?
NO
What is statistical significance denoted by?
Alpha
What is statistical significance?
The probability of incorrectly rejecting the null hypothesis when it is true
(the probability of making a type 1 error)
What is statistical power denoted as?
1- beta
What are type 1 and type 2 errors denoted by?
Type 1= alpha
Type 2= beta
What is statistical power?
The probability of rejecting the null hypothesis correctly when it is false
(the probability of not making a type 2 error)
What can increase the rate at which statistical signifance is achieved?
- By increasing the sample size
- By having a smaller standard deviation
What can be taken from trials that did not achieve statistical significance?
It does not mean there is not a relationship, just that there was not enough evidence to prove that a relationship exists
What is the indepandant samples t-test?
Used when want to compare the means of two groups
What criteria needs to be met before the independant samples t-test can be used?
- There is a random, unbiased sample
- Data being compared must be continuous
- Sample must follow normal distribution or have a large enough sample size
- Need evidence to prove equality of variances
How many individuals need to be in a group in order to ignore proving normality?
At least 30 in EACH group
How do you prove normality in samples?
Q-Q plot
Scatterplot of sample quantiles against expected quantiles
If 45* line fits data well= normaility
When would you need to prove normailty?
When there is < 30 people in each group of the study
How do you test for homogenity/equality of variances?
Levene’s test
What is needed in Levene’s test in order to use the independant sample t test?
P value > 0.05
Need to find enough evidence to not reject null hypothesis
What is the null hypothesis for Levene’s test?
Variances for group 1 = variances for group 2
What test is used when there is no evidence of equality of variabilty?
Welch’s t-test
What can be concluded if the confidence interval contains both negative and positive values?
Cannot decide with certainty which group’s mean is higher
Contains 0- not enough evidence to reject null hypothesis and p will be > 0.05
How can you decrease the confidence interval?
Increase the sample size
Decrease variability between groups
What is applied when conducting the independant samples t test on groups with at least 30 people?
Central limit theorum- A sufficiency large sample size’s distribution of means approximates to normaility
What is parametric and non parametric data?
Parametric= normally distributed
Non-parametric= does not follow a normal distribution
What test is used instead of the independant sample t-test for non-parametric data?
Mann-Whitney U test
What are cells, observed frequency, expected frequency and marginal total in contingency tables?
Cells= boxes that contain frequencies
Observed frequency= numbers outside bracket
Expected frequency= numbers in the bracket
Marginal total= row and column totals
What are the two tests used to analyse contingency tables?
Chi squared test
Fishers exact test
What does a greater chi squared result mean?
Greater overal discrepancies between observed and expected frequencies
Therefore more evidence to reject the Ho
What is the continuity correction that can be applied to chi squared tests?
Yates correction
What does Yate’s correction do to the p value?
Reduces value of numerator and therefore the value of chi squared so the p value will rise
What is the criticism of Yate’s correction?
It is too conservative, risk of false negatives (type II errors)
How do you know when to use Chi squared test v Fishers exact test?
If 80% of counts in contingency table are of size 5 or more= chi squared
If not = Fishers exact test
What is an odds ratio?
Statistic that quantifies the strength of association between two events, A and B
How are odds ratio calculated?
Why might odds ratio not always be accurate?
The size of the confidence interval will effect it