Stats Flashcards
Define bias
Systematic difference between the observed effect and the true effect. Not due to chance.
Define categorical data
categorical data refers to data is qualitative, describes a quality. It can either be ordinal or nominal
Define evidence based medicine
The use of the best available scientific research to guide clinical decision making
Define NNT
Number needed to treat refers to the number of patients needed to be treated with an intervention for 1 patient to have a benefit. Equals 1/ ARR
Define PPV and NPV
PPV: given the test is positve, what proppriton actually have disease= TP/ TP+ FP NPV: given that the test is negative, which porpotion don’t have disease: TN/ FN + TN
Define quantitative data
Quantitative data is numerical. It can be continuous or discrete. Continuous data can either be ratio (a zero point exists) or interval
Define relative risk reduction; absolute risk reduction
RRR= differnece between the incidence in the unexposed verus exposed / the incidnece in the unepxosed ARR: difference in the incidence between the unexposed and exposed.
Define risk ratio; define odds ratio;
risk ratio= relative risk: incidence in the exposed/ incidence in the unexposed.Used for RCT/cohort studies Odds ratio: odds of event in the exposed/ odds of the event in unexposed. Use in case control studies when can not determine incidence. Approximates relative risk, especially for rare events.
Define standard deviation; define variance
standard deviation: square root of the variance; vairance: the average of the squared differences from the mean = E (x- average x)** / (n-1)
define standard error of the mean (SEM)
SEM is the standard deviation of the sampling distribution. SEM= s/sqr(n) (sqr=square root)
Define Standardised Mortality ratio
Defined as the observed mortality rate/ expected mortality rate (as estimated from scoring systems such as APACHE II)
List non- parametric tests
Chi squared Wilcoxon Mann Whitney U test Kruskal wallis
What are the features of a normally distributed population
mean= mode= median bell shaped curve 68% of poupulation within 1 SD of mean 95% of populatio within 2 SD of mean
What are the parametric tests
t-test ANOVA
What are the Wilcoxon rank sum test, Wilcoxon paired sample test; Mann Whitney U test and Krsukal Wallis tests?
Wilcoxon rank sume test: non parametric equivalent of t test Wilcoxon paired sample test: non paramertric equivlanet of paited t test Mann Whitney U test: Non parametric equivalnet of t test Kruskal Wallist test: non parametric equivalent of ANOVA (i.e. more than 3 groups with continuous variables)
What assumptions are made in parametric tests
The data is continuous. The samples have the same variance and have been taken from a normally distributed population
What factors affect the power of a study
- The sample size 2. The effect size 3. the variation in the study population for the effect 4. the significance level chosen (alpha)
what is a meta analysis
a mathematical process which seeks to provide a pooled estimate of effect from several trials
What is a systematic review
a planned unbiased summary of evidence adressing a specific clinical question
what is ANOVA
ANOVA- analysis of variance - a paramtric test which determines if there is a difference amongst 3 or more groups with continous vairbales. Does not tell you which group is differnt.
What is statisitical Power
measures the chance of detecting a significant difference when such a difference exists. Equals 1- type II erorr. Usually want it to be at least >80%
What is the chi squared test
tests whether there is a real differnce in the frequency of categerocial events between two or more groups
What is the confidence interval
CI is the range around a sample mean within which you predict the true mean lies
What is the null hypothesis
states that there is no effect.
What is the P value
The probability of of the observed difference having occured by chance. Therefore give the likelihood of the Null Hypothesis being true.If P large: can not reject the null hypothesis. If P small: reject null hypothesis , knowing there is a smallchance that it is true.
What is the Receptor Operator Characteristic curve
Plots sensitvity (y axis) against 1-specificty (equivalent to the false positive rate) (x axis)
What is type 1 and type 2 error
Type 1: conclude there is a difference, when there isn’t. False positives. Determined by the significance level set Type 2 error: conclude that there isn’t a difference when there is on.e Flase negatives. Commonly caused by small sample size.