Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
71 Cards in this Set
- Front
- Back
4 Biases that can threaten the validity of a study:
|
1. Selection bias
2. Measurement bias 3. Confounder bias 4. Simply chance |
|
What is an event rate?
|
Incidence
|
|
CER
|
control event rate
|
|
EER
|
experimental event rate
|
|
How do you calculate EER in an RCT?
|
# of exposed w/ disease
----------------------- total exposed |
|
How do you calculate CER in an RCT?
|
# of nonexposed w/ disease
-------------------------- total control (nonexposed) (convert denom to 10,000) |
|
How do you calculate RR?
|
Relative Risk = EER/CER
|
|
How do you calculate the percent increase in risk?
|
By calculating the Relative Risk Difference
|
|
How to calculate Relative Risk Difference:
|
ARD/CER
|
|
What is ARD?
|
Absolute Risk Difference
|
|
How to calculate Absolute Risk Difference:
|
EER - CER
|
|
How do you calculate NNT?
|
number needed to treat = 1/ARD
|
|
What is the NNT?
|
The number of patients who would need to recieve the intervention in order for one event to occur.
|
|
What does the NNT allow you to do?
|
Compare outcomes and treatments directly
|
|
What is the stipulation for RCT to be able to generate a Relative Risk Ratio?
|
The outcome must be dichotomous
|
|
What do you do if the variable outcome is not dichotomous, ie is continuous?
|
Look at the mean change in the variable and compare mean differences.
|
|
3 measures of central tendency:
|
-Mean
-Median -Mode |
|
Mean
|
average
|
|
Median
|
middle value
|
|
Mode
|
most common value
|
|
2 measures of dispersion:
|
-Standard deviation
-Range |
|
Standard deviation:
|
variation around the mean
|
|
Range:
|
highest to lowest value
|
|
What is the measure of effect for continuous variables?
|
Mean difference
|
|
What are the measures of effect for Dichotomous variables (4)?
|
1. Relative Risk
2. Risk difference 3. Odds ratio 4. Hazard ratio |
|
How should you always err in clinical research?
|
On the conservative side - the cause is innocent until proven guilty for causing the effect.
|
|
What is the difference between Sample and Population?
|
Sample is a subset of the population studied, the latter consisting of everyone in whom you're interested.
|
|
N:
|
the number of persons in the sample. can't believe you made this card.
|
|
When formulating a research question, what do we begin with?
|
-Null hypothesis
-Alternative hypothesis |
|
What is the null hypothesis?
|
That there is no difference between exposed/unexposed groups. (presume innocence)
|
|
What is the Alternative Hypothesis?
|
That there really is a difference.
|
|
What are the four possibilities for your hypotheses at the end of a study?
|
1. Accept the Null and it's true
2. Accept the alternative and it's true 3. Accept the alternative but it's false 4. Accept the null but it's false |
|
What type of error is it when you accept the alternative and it's false?
|
Type I
|
|
What type of error is it when you accept the null and it's false?
|
Type II
|
|
Which type of error is worse; Type I or Type II? Why?
|
Type I - because you say there is a difference in exposed vs unexposed, caused by the factor, but it's not really true - totally not erring on the conservative side.
|
|
What is "alpha"?
|
The threshold of reasonable doubt we will accept to make a type I error
|
|
What is the usual value for alpha?
|
0.05 (5%)
|
|
What happens if alpha is greater than 5%?
|
The results are not statistically significant.
|
|
What is the difference between a Statistic and Parameter?
|
Parameter - a characteristic of the population. (true mean)
Statistic - a characteristic of the sample (SD, Event Rate (CI) |
|
What do we hope to do in RCTs?
|
Generalize the statistics generated by the sample, to describe the parameters of the population of interest.
|
|
What would a Type II error likely be due to?
|
Small sample
|
|
What is Beta?
|
The probability of a type II error.
|
|
What is the usual value of Beta?
|
0.20 (20%)
|
|
What is Power?
|
1-Beta = .80 (80%)
|
|
What does Power refer to?
|
The power of a study to detect a difference.
|
|
When flipping a fair coin, what are the odds that you would get heads 10 times in a row, randomly?
|
Less than 0.1%
|
|
What is the null hypothesis here? Alternative?
|
Ho = 10 heads would land in a row, randomly
Ha = 10 heads would land in a row not randomly, but due to some external influence. |
|
Since the probability (p) of getting heads 10X in a row randomly is soooo low, what is the conclusion?
|
Reject Ho and Accept Ha.
|
|
If a new drug lowers BP by 10 mm Hg lower than the old drug, how do you tell if the new drug truly is better?
|
By doing a statistical test to tell if the probability (p) of getting this difference by random chance alone is <0.05
|
|
How do you know if the new drug is better?
|
If the p value is low - it's not probalby just random luck that the new drug lower BP by 10mmHg.
|
|
4 Things necessary to calculate sample size:
|
1. Set alpha/beta ahead of time
2. Choose statistical tests 3. Make assumptions about what you expect - effect size, variability 4. Calculate how many subjects would be needed to recruit these levels |
|
What MUST BE in order for results to be significant?
|
P must be less than alpha!!!
(<0.05) |
|
What CAN'T you do with P's?
|
Compare P's from 2 studies - say one study is more significant than another because the P value was lower..
|
|
P
|
Probability of finding due to chance alone is small (alpha)
|
|
Power
|
Power to prove a relationship does not exist (1-beta)
|
|
Power is highly dependent on:
|
sample size
|
|
What is a Confidence Interval?
|
A description of statistical significance that gives more information on the size and accuracy of the effect measure than P does.
|
|
What makes the best CI?
|
-Narrower
-Doesn't cross null |
|
What does it mean if CI crosses the null?
|
It is not significant.
|
|
What is the null for mean differences (blood pressure)?
|
0
|
|
What is the mean for relative risks? (dichotomous variables)
|
1
|
|
What is the CI calculated from?
|
The mean
|
|
Formula for the 95% CI
|
mean +/- 1.96
|
|
4 Components in the PICO model for RCT studies:
|
-Population
-Intervention -Comparison -Outcome |
|
What does the RCT begin with?
|
The hypothesis that includes the PICO format
|
|
2 types of validity that we look for in an RCT:
|
-Internal
-External |
|
What is internal validity?
|
That the study results are free from bias and error.
|
|
What is external validity?
|
That the study results can be generalized - sample results attributable to the population.
|
|
How should an RCT investigator assign subjects to control vs expirimental groups?
|
RANDOMLY according to number sequences.
|
|
Why is randomization important?
|
It limits the effects of confounders.
|
|
What is "intention-to-treat"?
|
Keeping subjects in the group (case vs control) to which they were randomized at the beginning of the trial.
|