Important Definitions Flashcards
What makes a good hypothesis?
Valid data gathering and no fishing expeditions
What are the four steps of hypothesis testing?
- Make hypothesis
- Set criteria for decision
- Gather data and conduct statistical analysis
- Make decisions -> accept H0 or reject HA
What is the Null Hypothesis (H0)?
No difference
- All samples from same population
- Observed difference due to chance (random sampling)
What is the Alternative Hypothesis?
At least one sample from a different population
- Difference not due to chance
All statistics are based on…
Probability
What is our main goal in hypothesis testing?
Rejecting the null
If alpha < our result, then…
We accept the null
If alpha > our result, then…
We reject the null
What does mutually exclusive indicate?
Only one event can occur
Ex: heads or tails
What does independent variables indicate?
Either event could occur
Ex: roll a 6 on one die and a 6 on the other die
When do we use the additive rule?
With a mutually exclusive event
When do we use the multiplicative rule?
With an independent probability
What is sampling?
Selection of study subjects who will be
measured on some parameter to provide information about population
What are the representative sampling techniques?
- Simple random sampling
- Systematic sampling
- Stratified sampling
- Convenience sampling
- Multistage sampling
Simple random sampling
Each individual has equal chance of being selected
Ex: drawn out of hat
Systematic sampling
Select at regular intervals through an ordered list
Ex: interview 3rd patient for one week
Stratified sampling
– Separate population into strata based on a Separate population into strata based on a characteristic
–– Randomly select proportion of participants from each stratum
Ex: picking only freshmen, and then select randomly within freshmen
Cluster sampling
Population already in strata
* Randomly select groups of strata
Convenience sampling
Cold call
* Does not typically represent population
Multistage sampling
Combines more than one sampling
Ex: randomly selected PCPs chosen for survey, when answering, second detailed survey sent out to randomly selected sample
What is a qualitative variable?
Catergorical: groups, cohorts
Ex: ethnicity, gender, disease status, age group, stage of cancer, pain rating
What is a quantiative variable?
Continious: measured quantities
Ex: height, weight, temperature, blood pressure, temperature
What is an independent variable?
Input, manipulative
What is an dependent variable?
Outcome, response, predicted
What are measures of central tendency?
Mean, median, mode
What are measures of dispersion?
Range, interquartile range, percentiles
Standard deviations (SEM), max/min, kurtosis, skewness
What does sample size mean?
Number of samples in each group
Larger sample sizes tend to have more statistical power
How can we reduce variance?
Increase the sample size
Better define our sample populations
Censor outliers
Unimodal distributions
Normal and skewed distribution
What is the bell-shaped curve?
Gaussian Distribution
What is Gaussian Distribution used for?
Continuous variables
What is described of in Gaussian Distribution?
Mean, and standard deviation
In a skewed distribution…
Mean not equal to median
Positive skew
Mean > median > mode
Negative skew
Mean < median < mode
What is kurtosis?
Peakedness or flatness of frequency distribution
Higher kurtosis means…
Variance from infrequent extreme deviations
What is platykurtic?
Negative kurtosis
What is leptokurtic?
Positive kurtosis
What is a type I (alpha) error?
Probability of rejecting null when it is true
What is type II (beta) error?
Probability of accepting null when it is false
What are the rules of categorical data?
Data are arranged into catergories
Data are an independent, random sample from the population
Must be positive integer, > 0
What do we assume in categorical data?
Random sample of the population
What are the type of categorical data studies?
- Cross-sectional: recruit study group and define (at
least) two categories. - Prospective: recruit groups with some difference
(risk factors) and follow them. - Retrospective: start with condition and look
backwards at potential causes. - Experimental: recruit group, manipulate variables
- Diagnostic Test: positive + negative results
What are the requirements of the student’s t-test?
- Data is continuous (not categories)
- Data is a random sample from the population
What are the assumptions of the student’s t-test?
- Each of the 2 populations being compared is
normally distributed - Each of the 2 populations should have the same
variance
One sample t-test
Compare a group’s mean to a “benchmark” group
Independent t-test
Compare two independent groups
Dependent t-test
Compare 2 groups that are somehow related
Two-tailed t-test?
LECOM students will score significantly different than group X
One-tailed t-test?
LECOM students will score significantly higher than group X
ANOVA stands for…
ANalysis Of VAriance
What is ANOVA used for?
Comparing means of 3 or more groups
What are the assumptions of ANOVA?
Independent observations
Normally distributed data
Homogeneity of variance [Levene’s Test]
What is a post hoc test identify?
Specific groups that are different
When do you use Tukey post hoc?
Compare each group to every other group
When do you use Dunnett’s post hoc?
Compare all groups to a control group
When can you develop a correlation?
Both dependent and independent variables are continious
What does correlation not imply?
Causation
What are the detailed criteria for assessing evidence?
– Strength
–– Consistency
–– Specificity
–– Temporality
–– Biological gradient
–– Plausibility
–– Coherence
–– Experiment
–– Analogy
Strength in Bradford-Hill Criteria is defined as…
Larger association is more likely causal effect
Consistency in Bradford-Hill Criteria is defined as…
Stronger association when same findings observed by
different people and different samples
Specificity in Bradford-Hill Criteria is defined as…
Specific population (or risk factor) with specific disease
and no other likely explanation
Temporality in Bradford-Hill Criteria is defined as…
Effect happens after the cause
Biological gradient in Bradford-Hill Criteria is defined as…
More exposure = greater incidence
Plausibility in Bradford-Hill Criteria is defined as…
Plausible mechanism between cause and effect helpful
Coherence in Bradford-Hill Criteria is defined as…
Link between epidemiological and laboratory findings
increases likelihood
Experiment in Bradford-Hill Criteria is defined as…
“Occasionally it is possible to appeal to experimental
evidence”
Analogy in Bradford-Hill Criteria is defined as…
Effect of similar factors may be considered
Degree of correlation measured by…
r
…is coefficient of determination
Rsquared
What is Spearman’s Rank?
Nonparametric test for relationship between 2 variables
When do we use survival analysis?
Used to analyze data in which time until the event is of interest
Types of bias
- Selection bias (sampling)
** Measurement bias
–– Hawthorne effect
** Experimenter bias
–– Pygmalion effect
** Lead-time bias
** Recall bias
** Late-look bias
** Confounding bias
** Design bias
What is selection bias?
Sample is not representative of population
What is measurement bias?
Leading question
What is the Hawthorne effect?
Participant behavior altered by knowing
they are being studied
Experimenter bias (Pygmalion effect)
Expected results accidently communicated to
participants
Lead-time bias
False estimate of survival rates
Recall bias
Participants do not accurately recall events
Late-look bias
Participants with severe disease less likely to be
identified because they die first
Confounding bias
A factor in a study is related to co-factor being
examined
Design bias
Choosing inappropriate study design to answer
question
What is incidence?
Measure new cases of disease or injury in of disease or injury in population over specified time period
What is prevalence?
Measure of total number of cases of disease or injury in population over specified time period
What is validitiy?
Test detects what it was designed to test
What is test sensitivity?
Proportion of pts with disease where screening test
is positive
Higher sensitivity, better test in finding diseased
pts
What is test specificity?
Proportion of pts without disease where screening
test is negative
Higher specificity, healthy pts not diagnosed as
diseased
Case-control study
Find cases and look in the past
Cohort study
Choose a group and follow
What are the advantages/disadvantages?
- AdvantagesAdvantages
–– Can estimate incidences
–– Can calculate risks
–– Exposure precedes disease
** DisadvantagesDisadvantages
–– Rare disease
–– Slow time from exposure to disease development
–– Loss to follow-up
–– More expensive – longitudinal in time
What is phase I in clinical trials
Small number of healthy volunteers
What is phase II in clinical trials?
Small # of pts with disease
What is phase III in clinical trials?
Randomized Control Trial (next)
What is phase IV in clinical trials?
Surveillance after approval and use
What are key elements to RCT?
Treatments, outcomes, benefits
What is meta-analysis?
A quantitative, formal, epidemiological study design used to
systematically assess previous research studies to derive
conclusions about that body of research
What are the advantages of meta-analysis?
Improve precision, answer questions not posed by individuals studies, settle controversies arising from apparently conflicting studies
What are the principles of meta-analysis?
- It is important to clearly state your hypotheses and
search strategies to decrease bias and heterogeneity - Clearly identify patient population, inclusion and
exclusion criteria, desired outcomes and intervention
to be studied - Make sure you are comparing apples to apples!