Stats Flashcards
what is cohens Kappa statistic?
measure of agreement between raters for categorical measures
there are two different types of agreement dependant on which data types?
continous
categorical
equivalent of kappas for continous data?
Intraclass correlation (ICC)
is agreement the same as correlation?
no
which plot can be used to calculate the means and differences between two techniques?
Bland-Altman method
Kappa and weighteed kappa is used for which type of data set?
categorical ratings
ICC and bland altman is used for which type of data set?
continuous
when are repeated measures seen?
measuring same experimental unit on several occasions
give examples of repeated measures?
glucose uptake by cells at different time points
cholesterol in RCT of new statin at 3,6 and 12 months
what test is commonly used when analysing repeated measures?
t-tests
multiple testing when analysing repeated measureshas what effect on error?
increases type I error
how is the inc in type I error with multiple testing accounted for?
adjustign p-value
Bonferoni correction
why is bonferoni correction necessary?
multiple tests assumes aim of study is to show sig difference at every time point but most studies aim to show overall effect
how would you describe the primary hypothesis when dealing with overall analysis of repeated measures?
primary hypothesis os global
what is ANOVA?
Analysis of Variance
determines whether there are any statistically significant differences between the means of 3 or more unrelated groups
what is sphericity?
refers to teh issue of similarity of variances in the differences between treatments
when sphericity is not met what can be used as an alternative to the correction factors?
MANOVA
has less power than repeated measures analysis so generally avoided
what can be used in addition to MANOVA if spherivity is not met?
Greenhouse-Geisser
Huynh-Feldt adjustment
what are parametric tests?
involve estimating parameters such as the mean and assumes that distribution of sample means are normally ditributed
what is skewness?
skewed distributions i.e. postive or negative
if distribution is skewed what can be done to normalise data?
use log
what are non-parametric tests?
tests for situations where nothign is assumed about the distribution
Non-parametric tests are sometimes called?
Distribution free tests
what is a practical difference between parametric and non-parametric methods?
NP mehtods use ranks of values rather than actual values
what is the median?
value above and below which 50% of the data lie
in symmetric distributions the median and the mean are the same true/false?
true
in skewed distributions which is more appropriate the median or the mean?
the median
what is a T-test?
used to test whether the mean of a sample is significantly different from another
relies on sample being drawn from a normally distributed population
if a sample is not normal what can be used instead of a t-test?
Wilcoxin Signed Rank Test
what is the difference between teh Wilcoxin signed rank and the Wilcoxin Rank sum?
Signed Rank= paired t-test
Rank Sum= independant t-test
give an advantage and a disadvantage of a paired t-test?
advantage= uses all detail of available data- most powerful test if data normally distributed
disadvantage= assumes data are a random sample from a population which is norally distributed
describe paired and not paired comparisons
paired- same sample measured on 2 diff occasions
not-paired- two independant samples
(‘matched’ samples by age/gender are paired)
what is another name for the Wilcoxin Rank Sum?
Mann-Whitney test
when is the Mann whitney test used?
used when you want to compare two unrelated or INDEPENDANT groups
what assumptions does the Mann Whitney have?
distributions have similar shape but need not be normal
can be used for small/large sample sizes
what is the spearman rank correlation?
tests monotonic relationship between 2 measured variables
lies between -1 to +1
a spearman rank correlation close to 0 is strong/weak?
weak or no correlation
when is Chi-squared testing used?
when comparing 2 or more groups of categorical or nominal data (as opposed ot measured data)
when comparing 3 or more groups and data is not normal what is used when independant and related samples?
independant sample= Kruskal-Wallis
related samples= Friedman
what are survival regression models?
risk of event (mortality) in next time interval t, conditional to having survued to start of interval t
(expressed in terms of hazard function)
what is hazard?
hazard rate is an instantaneous rate of events as a function of time
which model expresses the relationship between the hazard and a set of variables or covariates?
The cox model
what is hazard ratio?
ratio of hazards (HR) in two groups e.g. men v women
(it is the improvement of one group over the other)
what happens to the HR over time?
stays constant
explain the following hazard ratios: HR= 1, HR= <1, HR= >1
HR= 1 do not reject null hypothesis (i.e. no difference)
HR= <1 reduction in hazard relative ot comparator i.e 0.6 is 40% reduction
HR= >1 increase in hazard relative to comparator i.e. 1.7 is 70% increase
what is the logrank test?
is it parametric or nonparametric
a hy[pothesis test used to compare survival distributions of two samples
if a 95% CI for a HR included 1 can you reject the null hypothesis?
no
how is hazard ratio interpretated for continuous and categorical variables?
continous i.e. age- HR represents inc per unit in age
categorical i.e. dukes staging- HR represents inc per category
when is multiple regression modelling used?
to asses relationship between two variables whilst allowing for another variable
what is a cofounder?
factor which is related to both the variable of interest and the outcome
give two examples of selection criteria used when model has many variables?
Alaike’s Information Criterion (AIC)
Schwarts’s Bayesian Information Criterion (BIC)
multipe regression models are the most used analytical tool in which type of research?
quantitive research
define correlation
strength of the linear relationshop between two variables
what is regression analysis?
determines nature of a relationship i.e. is there a relationship between alcohol units drank and liver cirrhosis
what is pearsons coefficient of correlation?
measures the strength of the linear relationship between one dependant and one independant variable
what are the values of pearsons coefficient?
r= +1 positive correlation
r= -1 negative correlation
r= 0 no linear relationship
define residual
a residual is teh difference betwween the predicted value (fitted line) and the actual value or unexplained variation
what is a durbin watson test?
test for autocorrelation in the residuals from a statistical regression analysis