Critical numbers Flashcards
What is a sample?
A selection from a population which aims to represent the whole population.
Name some types of bias and explain
Sampling bias, some people are more likely to be included in your sample than others. (omission, inclusive bias)
Recall bias, people cannot remember information correctly
Social- desirability bias, change answers to more acceptable ones
Information bias, errors in your measurements.
What is a confounding factor?
something that is related to the outcome and the characteristic of interest
What are the categories of studies?
– Experimental vs. Observational
– Retrospective vs. Prospective
– Individual vs. Population level
What is an experimental research method?
One where the researcher has made some kind of intervention eg crossover trial or RCT
What is an observational study?
There is no intervention data is just collected about what happens, E.g. case-control, cross-sectional,cohort,ecological studies
What is a retrospective study?
One which looks back at what has already happened case-control
What is a prospective study?
Collect information then follow up over time Cohort study
What is an individual study?
Collect information about individuals all studies except ecological.
What is a popilation study?
Talk about a whole population
What is the ecological fallacy
making inferences from populations about an individual.
Describe case-control studies
Find individuals with the outcome and a similar group without and take a random sample of each and see who had the eposure compared to others.
What are the strengths and weaknesses of case-control studies?
Strengths- good for rare diseases as sample can be focussed, quick results as its retrospective, can look at multiple causes, it is inexpensive.
Weaknesses- subject to recall bias, can’t prove causality, prone to selection and information bias, hard to choose controls, incidence of disease cannot be calculated
Describe a cross-sectional study
takes a snapshot of what is happening at the current time
What are the strengths and weaknesses of a cross-sectional study?
Strengths- quick to do, relatively cheap to generate hypotheses, few eithical problems
Weaknesses- could be a medical oddity, prone to sampling bisas no time reference
Describe a cohort study
Collect information on a sample and follow- up over time to explore who gets the outcome
What are the strengths and weaknesses of a cohort study?
They distiguich anecedent causes from concurrent associated factors. can calculate relative and absolute risks. can look at many outcomes or exposures less chance of bias as measured before disease onset/
Weaknesses- cannot be certain of causation, long study periods, follow up can be a problem, diagnosis of disease may change over time.
Describe a RCT?
Have multitple groups with different exposures compare the outcomes to get a causal relationship.
What are the strengths and weaknesses of an RCT?
Strengths- should equalise confounding factor effects, randomly allocating reduces bias blindign can be used, significance tests are better, the confounders and mabiases are minimised
Weaknesses need a big enough sample to show it works, volunteer bias can be a problem, ethical issues can be a problem for studies and poor compliance can lead to loss in power of the study.
What is a crossover trial?
an extension to an RCT. everyone does all the arms of the study. which reduces confounding even more as each person can be compared across the arms. thre can be carry-over effects and more technical analyses
What steps should be taken in an RCT to minimise bias?
Blinding, randomisation, placebos, matching
What are the two main groupings for variables?
Categoric and numeric
What are the types of categoric variables?
Binary, ordinal, nominal
what are the numeric variables?
Discrete and continuous
What is the odds?
number with the outcome/ number without the outcome
How can you quantify differences?
Risk differences, risk ratio and risk difference
What is risk difference?
the difference between the two risks you have calculated
What is a risk ratio?
divide one risk by the other. the top group is the focus group compared to the other one.
How do you interpret a risk ratio?
RR> 1 the focus risk is higher
RR=1 the two groups are the same
RR<1 the focus risk is lower than the other
1 is no difference
How can you swap the focus of the risk ratio is?
inverse 1 divide by it
What is odds ratio?
Odds divided by odds
How do you interpret odds ratio?
OR> 1 the focus risk is higher
OR=1 the two groups are the same
OR<1 the focus risk is lower than the other
1 is no difference
Why might you use risk ratio?
It puts it in context more
Why do we use Odds ratios?
they are useful for some statistical methods
If something is very rare how does OR and RR compare?
RRroughly= to OR for rare outcoumes
If something is more prevalent what happens to RR and OR?
RR<
What is the median
middle value in sequential order
What is positive skew?
where the peak is to the low end most of the values are in the lower end, the mean is greater than the median
What is negative skew?
most of the values are to the right, the mean is less than the median
How to decide which measure of spread to use?
if it is symmetric then use mean and SD otherwise use median and IQR as they are less affected by skew
What is the use of the normal distribution?
The sd can tell you about percentage certainty
What are the limits of correlation coeffiecient?
-1 perfect negative correlation 0 and +1
What is standard error?
How well your sample representing the population.
How can standard error be reduced?
Enlarging the sample size the more similar the people are.
What is the formula for the standard error of a mean?
SD/root(n)
What is the difference between standard error and standard deviation?
Standard deviation is a descriptive value about the data collected while standard error is an inferential number about how well our estimate represents the true population value
What is the use of confidence interval?
It is often used as a comparative value between data sets. can be used for inferential statistics
What are confidence intervals?
The true value is quite certain to lie between those two points.
What are confidence interval calculated from?
Standard error and SD values
What is the null hypothesis?
There is no link between the two variables
What is a p value?
Probability that the mean could be from the standard deviation.
If mean is close to the null what will the p value be?
Close to one
How should you phrase rejecting the null hypothesis?
The evidence suggests to reject the null hypothesis
What is the generally accepted significant p value?
p=0.05 for statistical significance
What is another significance test?
One sample t test, two sample t test, chisquare tests, ANOVA test, Pearson correlation coefficient
What is regression?
Plotting the correlation between variables using y=a+bx
What is the effect of using multivariable method?
It accounts for the effect of confounding factors
How can you appraise study design?
Who is sturied? are there missing groups over sampling? is it clear what the aim is
What can you appraise the descriptive statistics?
Summariesed data appropriately, Normal distribution, SD
What can you appraise the inferential statistics?
p values CI did they look at normality test
