medical statistics Flashcards
what is a statistic
numerical summary of a sample
what is a parameter
numerical summary of the population (use sample stats to estimate parameter values)
types of variables
categorical
numerical
time to event
categorical variables
nominal
ordinal
numerical variables
discrete
continuous
nominal variables
categories are mutually exclusive and unordered
e.g.
- blood group
- ethnicity
- country of birth
ordinal variables
categories are mutually exclusive and ordered
e.g.
- disease stage
- education level
- heart murmur grade
discrete variables
take only integer values (0,1,2…)
e.g.
- number of meds
- number of hospital admissions
continuous variables
take any value in a given interval
e.g.
- age
- weight
- blood pressure
standard deviation graphical values
1 s.d. = 68%
2 s.d. = 95%
3 s.d. = 99.7%
z score =
(observation - mean) / (standard deviation)
steps in hypothesis testing
- state the null and alternate hypothesis
- decide the appropriate statistical test
- use the test to calculate the p-value
- weight the evidence against the null
if p value is smaller than or equal to 0.05…
statistically significant, unlikely due to chance so sufficient evidence to reject the Ho
if p value is larger than 0.05…
not statistically significant, result likely due to chance so insufficient evidence to reject the Ho
limitations of the p value
- mixes strength of an accosication with the sample size (precision)
- p value goes down as sample size increases
bivariate tests
group 1 - two numerical variables
group 2 - two categorial variables
group 3 - one categorical, one numerical
what statistical tests are appropriate for two numerical variables
correlation/regression
- correlation (two-sided association)
- simple linear regression (one-sided association)
what statistical tests are appropriate for two categorical values
chi-squared tests
- chi-squared test (unpaired)
- mcnemar test (paired)
what statistical tests are appropriate for one categorical and one numerical variable
- 2 groups = t-test (paired and unpaired)
- > 2 groups = ANOVA (unpaired) & ANOVA for repeated measures (paired)
correlation coefficient (r)
measures strength of the linear association between two numerical variables
bateen -1 and +1
