statistical terminology Flashcards
Attributable risk
how much greater frequency of disease is in exposed group compared to non-exposed group
assuming exposure is casual
attributable risk =
incidence in exposed-incidence in non-exposed
excess fraction =
incidence in exposed/incidence in non-exposed
case
an individual with the outcome under study (have disease/illness/event)
not same as clinical definition
count
number of cases that occurred in a population
useful in planning services, not for comparison
chi squared test
statistical procedure fro determining if 2 proportions are similar
cohort study
study where individuals are chosen on basis of exposure and followed over a period of time to allow occurrence of the exposure of interest
produces relative risk
confidence interval
an estimated range of values calculated from a given set of sample data which are likely to contain the true population value
relative risk - cannot include 1
difference - 0?
confounding
possible explanation for study finding if confounding variables have not been taken into account
confounding variable
a factor associated with exposure and outcome of interest
control
person without the outcome of interest or not receiving the intervention
exposure
people come in contact with something hypothesised to have an effect on health - risk factors
incidence
number of new cases of outcome of interest occurring in a defined population over a defined period of time
can be interpreted as probability that the individual will develop the disease in a given time
matching
method for controlling for the effect of confounding at design stage of case control - controls have similar distribution of confounding factors to cases
normal distribution
set of values and frequencies that describe many things - symmetrical distribution is foundation of many statistical tests because if know sd and mean - can draw every point on the curve
null hypothesis
first stage in performing a statistical test
states there is no difference in the groups or that the rates are equal - relative risk = 1
odds ratio = 1 in case control study
rejected if the confidence interval doesn’t include the value expected under null ie 1 for RR or OR
for normally distributed groups - reject the null hypothesis if 0 - no difference between averages
odds
way to express probability
odds ratio =
odds of exposure in exposed/odds exposure in controls
odds in terms of probability =
probability/(1-probability)
odds ratio
for case control studies
ratio between 2 odds
good estimate of relative risk - which cant be measured because we don’t know incidence
1 - exposure no more likely in case or control
>1 exposure more likely in case - increase risk of disease
<1 exposure more likely in control - protective
outcome
event or main quantity of interest in a particular study
population attributable risk
population excess risk - measure of risk of outcome in study population attributable to the exposure of interest
population excess fraction
population attributable fraction
measure of proportion (fraction) of cases observed in study population attributable to exposure of interest
prevalence
number of cases of an outcome of interest in a defined population at a particular point in time - point prevalence
includes new and existing cases
p-value
probability of obtaining the study result if null hypothesis is true
smaller p value - easier to reject the null hypothesis
randomisation
a method for ensuring both groups in a clinical trial have similar proportions of confounding variables eg age
rate and risk
personal risk
population rate
both terms imply proportion
rate and risk =
number of people with outcome of interest/number of people at risk of outcome
regression
method for controlling effect of confounding at the analysis stage of te study
statistical modelling is used to control for omne or many of the outcomes
relative risk
measure of association between exposure and disease
ratio of incidence rate in exposed group and incidence rate in non-exposed group
1 = incidence same in exposed and not exposed - data shows no association between exposure and disease
>1 - positive association or increased risk among those exposed to the factor
<1 - inverse association or decreased risk among those exposed - protective
relative risk =
incidence in exposed/incidence in non-exposed group
restriction
control confounding at design stage of the study
sample
relatively small number of observations from which we try to describe the whole population from which the sample has been taken
find mean of sample and use confidence interval to describe the range within which we think the population lies
standardisation
control confounding at analysis stage
used to produce SMR
statistical test
only way to decide if results are due to chance or are real
take one value representing the observed difference between groups and compare against a table of appropriate mathematical distribution - eg normal distribution to see how extreme
subtract population mean from value and dividing by the population standard deviation - look up result in table of standard normal deviation (mean 0, sd 1) to find out what proportion of values are greater than this
stratification
method for controlling effect of confounding at analysis stage - risks are calculated separately for each category of confounding variable