68 terms

# Statistics

#### Terms in this set (...)

What is a null hypothesis (H0)?
States no difference present between groups
If a study is without statical significance in endpoint --->
Fails to reject null hypothesis
What is the null hypothesis in the Aim High Trail?
no difference in cardiovascular events (primary outcome) between using simvastatin alone and niacin combined with simvastatin.
What is the alternative hypothesis (H1)?
State a difference exists between groups
If study finds statistical significance -->
Reject the null hypothesis
What is the alternative hypothesis in the AIM High trails?
There is a difference between primary outcomes between simvastatin alone comported simvastatin combined with niacin
In a study why isn't the entire population studied?
It would take too long, too many people, costly. Key to good study is to have a good randomized sample that represents the population.
What is nominal data?
Categorical variables; no sense of ranking
What are some examples of nominal data?
Gender, mortalitiy, Yes/No type of questions
What is ordinal data?
Numbers used to indicate rank-order, without defining the magnitude between them.
What are some examples of ordinal data?
NYHA for heart failure, Pain Scale, Glascow Coma Scale
What is continuous data?
It can either be Interval or Ratio
Interval: Units of equal magnitude, rank-order, but it is without an absolute zero.
Rank: Same as interval, but there is an absolute zero
What is some examples of continuous interval data?
Temperature
What are some examples of continuous ratio data?
Pulse, blood pressure, age
Between mean, median and mode which is skewed by outliers?
Mean
True or false continuous data is always analyzed by parametric data if its normally distributed
True
What does parametric data assume?
Assumes that population studied or data set follows a normal distribution
What kind of data should the Chi-squared test be done?
Nominal --> independent with two groups
What kind of data is two sample t-test?
Continuous --> Independent with two groups
When is Mann-Whitney U used?
Ordinal --> Independent with two groups
When the nonparametric data is negative and skewed to the left what is the relationship between mean, median and mode?
Mean < Median < Mode
When the nonparametric data is positive and skewed to the right what is the relationship between mean, median and mode?
Mean > Median > Mode
What is interquartile Range?
Measure of variability directly related to median
• Difference between 25th and 75th percentile values
• Used for describing data that is nonparametric
What is range?
Spread between the lower and highest values
What is SD?
Variability of data around the mean
What are independent groups?
Not the same patients in each groups (patient in each group is different)
What are dependent groups?
Groups being studied are not different
(examples: patient in cross over study, identical twins, right v left eye in same patient)
What kind of data is mean age? (continuous, nominal, ordinal)
Continuous, --> ratio data
What if you are looking at age less then 65 or greater then 65? (continuous, nominal, ordinal)
Nominal
What is male v. female? (continuous, nominal, ordinal)
Nominal
What is history of myocardial infarction? (continuous, nominal, ordinal)
Nominal
What is glucose? (continuous, nominal, ordinal)
Continuous --> ratio
What is duration of prior statin therapy; <1yr, 1 to 5 yr, <5 yr? (continuous, nominal, ordinal)
Nominal, because you are putting them into three groups
Do you ever accept the alternative hypothesis?
No, you either reject the null hypothesis, or you fail to reject the null hypothesis
What is a type one error?
Its when you reject the null hypothesis but the null hypothesis is true..(Type I kills everyone)
What is a type II error?
Its when you accept the null hypothesis but the null hypothesis was false..(it makes the professors blue)
When is it ok to accept or reject the null hypothesis?
You make the correct decision when:
1) The null hypothesis is false and you reject the null hypothesis
or
2) The null hypothesis is true and you accept the null hypothesis.
What is the probability of committing a type I error?
Alpha is the probability of committing a type I error. Alpha corresponds to the p values (<0.05)
What is the probability of committing a type II error?
Beta
If you accept a false null hypothesis, you are making a:
• Type I error
• Type II error
• Type III error
• Type IV error
Type II error
What is power?

Power = 1-beta
The probability of discovering a true relationship (Probability that a statistical test can detect a significant different when it in fact truly exists).
What does it mean when the beta is set at 0.20 (power of 0.8)?
20% chance of missing an association and an 80% chance of finding an association.
When are power calculations used?
When designing a study to determine the sample size needed to find a difference
True or False:
Alpha value should be determined by a-priorir
True
What is a P value?
The boundary or threshold that separates the plausible and implausible is 5 times in 100 (p=0.05)
What does P=0.01 mean?
It means that there is a 1% chance that the results are due to chance.

(results that fall beyond P<0.05 are considered to be statistically significant)
What does statistical significant mean?
Result is sufficiently unlikely to be due to chance that we are ready to reject the null hypothesis
Does the statistical significance tell you the degree of significants?
No
Is a p<0.000001 more clinically significant than a P<0.01?
No
If a researcher sets p-value at 0.05, what is the probability of making a Type I error?
• 0.95
• 0.05
• 0.01
• 0.9
0.05
What does confidence interval mean?
Defines in internval likely to include the true population of data distributions.
What does 95% CI mean?
95% confident the population mean lies within this range
Risk calculators are mainly used for nominal outcomes what do they estimate?
Estimate the magnitude of association between exposure and outcome

* E=exposed
C=unexposed (control)
What is another term for incidence?
Absolute risk is another term for incidence
What is the equation for absolute risk (AR)?
What is Absolute Risk Reduction (ARR)?
ARR is the difference in event rate between studied group
What is correlation?
Strength of the relationship between 2 variables
What is the correlation coefficient (r)?
R: -1 to 1
• -1= perfect negative linear relationship
• + 1= perfect positive linear relationship
• zero= no relationship
What is r equal to in the H0 (null hypothesis?
There is no correlation
If r = 0.80 (95% CI 0.50-0.99), does this imply greater correlation?
Yes, there is a positive correlation because r is positive and 0.8 is greater then 0.5 then there is a gretater correlation.
Does correlation imply causation?
No, you need to look at what variables you are looking together. "r" does not have the ability to determine which came first"
What is survival analysis?
Studies the time between entry into a study and some event (any number of nominal endpoint; death, treatment response, etc.) patients are followed until they experience event or lost to follow-up
Why do survival analysis?
-The data is "censored"=event of interest (death) not observed in all patients or patients are lost to follow up
-Time to event data are usually skewed and analysis requires a nonparametric method.
What is the difference between Actuaria method and Kaplan-Meier?
• Actuarial Method- Endpoint assessed at predefined intervals
• Kaplan-Meier-Uses varying time intervals;Cumulative proportion surviving is recalculated every time an event occurs.
What is median survival time?
Time corresponding to an estimated probability of survival
What is hazard ratio?
The ratio of hazards is essentially "relative risk" but is done for survival studies
What does a HR >1 mean?
An HR > 1 suggests the group is more likely to experience the event
How do you interpret number needed to treat?
For every NNT patients with P, you will prevent one O if you treat
with I instead of C.