critical numbers Flashcards
experimental trials
researcher intervenes/ changes something
- randomised control trial
- cross-over trial
observational trials
researcher observes only
- case control
- cross-sectional
- cohort
- ecological
what uses retrospective data
case-control study
case-control trials are subject to which kind of bias
recall bias
what uses prospective data (collect data at start follows up over time)
cohort study
which does not use individual data
ecological
which uses a whole population
ecological
what is a case-control study
finds individuals with a specific outcome and some without
take random sample of each
look to see who had exposure
what is a cross-sectional study
snapshot of time/ whats happening now
who currently has exposure/ outcomes
what is a cohort study
collect information on a group of people and follow up over time
what is a randomised control trial
multiple groups
expose them to different conditions
what is a crossover trial
extension of RTC
everyone gets exposed to all conditions but in different orders
steps to avoid bias in an RTC
blinding
randomisation
placebos
matching
ordinal variable
2 or more categories with a natural ordering
nominal variable
2 or more categories with no natural ordering
discrete variable
where there is a distinct number of values e.g. age in years, number of times something happens
how to work out the odds of people who like cake
people who like cake/people who dont
for risk ratio a/b, what does a value higher than 1 mean?
risk in group a is larger than group b (if the number is 1.34, the risk in group a is 34% higher than the risk in group b)
an odds ratio of 1.9 means what (given odds a/ odds b)
90% increase in odds of an outcome for a compared to b
when does the risk ratio roughly equal the odds ratio
when the outcome is rare
when is the risk ratio much smaller than the odds ratio
when the outcome is more prevalent (common)
why can using the risk ratio in the media be misleading
it doesnt take into account sample size
what is the standard deviation
a measure of how spread out the values are from the mean
if the data is skewed, what should we use for the average
median and IQR
what should you not use for averages when the data is skewed
mean and standard deviation
what do we use to represent the standard deviation
σ
what do we use to represent the mean
µ
what is varience
σ^2
in pearsons correlation coefficiant (denoted r), what is a perfect positive correlation?
+1
the 2 values that are 1.96 σ away from the mean on each side represent how much of the data
95%
How do you work out standard error?
Standard deviation / root (n)
What is Standard error?
How precise the mean estimate is
What does CI stand for?
Confidence interval
How do you write that “we are 95% confident that the toe mean weight for cats is between 3 and 8kg”
95%. CI: 3kg to 8kg
What confidence interval is usually used and what is the exception
Normal is 95% but genetics uses 99%
How do you calculate the confidence interval?
Mean- ( 1.96 x SE ) = lower bound
Mean + ( 1.96 x SE ) = upper bound
How do you get from a 95% confidence interval to a 99%, confidence interval?
Multiply by the standard error (SE)
When can’t you use the SE?
When the data is not normally distributed and when n<20
As the population size increases, the standard error…
Decreases
A small p-value means what about the null hypothesis?
We reject the null hypothesis
What does a larger sample size do to the p-value?
Makes it smaller
If the null hypothesis is not in the 95% confidence value, the p-value will be…
p <0.05
What test is used to compare one mean vs a hypothesised mean?
One-sample t-test
What test is used to compare means between 2 groups?
Two-sample t-test (student’s t-test)
What test is used to compare 2 binary variables?
Chi-squared test
What test is used to compare means from 3t groups?
ANOVA test
What test is used to compare two numeric variables e.g height vs. weight
Pearson correlation coefficient
If the p value is 0.03 and the hypotherised value is 10, the confidence interval can’t be outside what?
7-13
What is Standard error?
How precise the mean estimate is
