
Pvalue:
probability of false positives in the study result

Null Hypothesis:
will be 0 for the difference in groups; while the null will be 1 for OR, HR, RR

Mean:
the average, this will be the middle of the bell shaped curve

Median:
the value that divides a series of numbers in half when they are listed in order, this will be used for skewed data that does not conform to the bell shaped curve

Mode:
the most frequently occurring number in a series

Standard deviation:
found using the patient population, square root of the variance, (the average sum squared difference from the mean); measurement of participant variability, measurement in the variability of data, how spread out it is

Interquartile Range:
the interquartile range (IQR), also called the midspread or middle fifty, is a measure of statistical dispersion, being equal to the difference between the third and first quartiles. IQR = Q3 − Q1

Skewed distribution:
nonnormal bell curve

Normal distribution:
normal bell curve, symmetric around the mean with the mean as the peak of the curve going out toward but never reaching 0

Standard error of the mean:
found with the clinical data, the spread of the sampled means for the data gathered, the measure of the precision/variability of the measurement (SD)

Confidence Interval:
“the neighborhood of the truth” An estimated range of values around a point estimate. Example: The 95% confidence interval says there is a 95% chance that the actual value found is within the C.I. Precision of study results, range of values that is likely to contain the population parameter

**Type 1 error is considered worse because….?
it would lead to unnecessary treatment of patients**

Type I error:
stating there is a difference in groups when there isn’t one , incorrectly rejecting the null hypothesis AKA failure to accept the null, will result in false positives

Type II error:
stating that there is no difference in groups where there is one, incorrectly accepting the null hypothesis AKA failure to reject the null, will result in false negatives

Power:
the probability to correctly reject the null hypothesis when you should. Mathematically defined as 1Type II error rate, depends on: sample size, difference between groups and type 1 error

Clinical significance:
Is the study result of practical interest? Do other findings matter?

Chisquare test:
comparison of categorical data for large sample chosen, may be used to compare groups…test whether observed frequencies are different from expected frequencies in a data table

Fisher's exact test:
categorical for small sample chosen (>5 subjects)

The ttest:
a statistical test used to detect the difference in two means, two groups, and factor in variability in data commonly used for continuous data, comparison of 2 different groups

The paired ttest:
a ttest for used when comparing two means that are within the same group Ex. The mean at the beginning of a study and the mean at the end of the study, comparison of dependent groups

Wilcoxon test:
a test for statistical significance of data that is not on a normal bell curve distribution (nonparametric), used for paired data–used for group comparisons, rank testing

MannWhitney U test:
a test for statistical significance of data that is not on a normal bell curve distribution (nonparametric), used for unpaired data –used for group comparisons, rank testing

ANOVA (analysis of variance):
A way to analyze groups of means to see if they are equivalent or not; if the ANOVA model fits the data well, and if a statistically significant difference is detected then posttesting is conducted

Posthoc testing:
 compare the group pairs, done in the second stage of statistical analysis…three types:
 • Tukeyused if the groups are unequal in size
 • Bonferronifor both equal and unequal groups
 • Scheffévery conservative to minimize type 1 error

Linear regression:
explains the differences in means. A calculation of the line of best fit passing through a set of data, which will allow for prediction about direction and amount variables change

Multiple linear regression:
explains the differences in means, in addition to explaining the differences in groups it can also be adjusted for age, gender, smoking, cancer etc…

Logistic regression:
allows for comparison of differences in odds between groups, results are an odds ratio which is a slight over estimate of relative risk

Multiple logistic regression:
in addition to explaining differences in OR between groups, they also adjust for age, gender, smoking, cancer etc…

Parametric tests:
ttest, ANOVA, regression

Nonparametric tests:
for ranking: Wilcoxon, MannWhitney; for categorical data: Chisquare, Fisher’s exact

Temporality:
cause must come before effect

Repeatability:
the effects must be repeatable

Biological Gradient:
the does response effect—small dose and small response v. big dose and bigger response

Reversibility:
dechallenge v. rechallenge aka the interventional effect

Plausibility:
Does what’s happening makes sense according to biological knowledge at the time



Numbers Needed to Treat (NNT):
the number of patients who would need to be treated in order to prevent one additional bad outcome

Numbers Needed to Harm (NNH):
the number of patients who would need to be treated in order for one bad outcome to occur

Absolute Risk (AR):
 •Mainly used with RCT
 •Probability of disease in the exposed group minus the probability of disease in the unexposed group
 •Represents the excess risk due to exposure to the factor under investigation

Interpretation of Relative Risks (RR), their confidence intervals, and the type of study that reports them:
 • RR is used to assess the influence of treatment/prevention strategies of potential hazards upon the prevalence of a given condition in a given population
 • RR will typically be used in cohort studies
 • RR is the probability of disease in the exposed group divided by the probability of disease in the unexposed group
 • RR >1 there is a positive association with the risk of disease
 • RR<1 there is a negative risk association with the risk of disease
 • RR=1 there is no association with the risk of disease

Interpretation of Odds Ratios (OR), their confidence intervals, and the type of study that reports them:
 • OR is used to assess how exposure to something effects disease
 • Used with case control studies
 • OR >1 those who are exposed are more likely to be diseased
 • OR<1 those who are exposed are less likely to be diseased
 • OR=1 there is no association with exposure and disease
 • If 1 is included within the confidence interval then the results are not considered statistically significant

Interpretation of Hazard Ratios (HR), their confidence intervals, and the type of study that reports them:
 • An estimation of harm given an exposure to a specific hazard
 • RR is the hazard of disease occurrence in the exposed group divided by the hazard of disease in the unexposed group over time (RR/time)
 • Used to assess the potential hazards of upon the nearness of an event
 • Used in studies looking for longer survivals due to a harmful or beneficial exposure
 • HR >1 there is a positive association
 • HR<1 there is a negative risk association
 • HR=1 there is no association

CrossSectional Design:
assess health status and exposure level of subjects at a point in time

Case Control:
retrospective observational study comparing diseased and nondiseased groups

Cohort:
prospective observational study comparing diseased and nondiseased groups

RCT:
prospective experimental design where the sample is broken into 2+ groups who are then put into categories such as treatment, placebo, alternative treatment, double dosage etc.

What Kind of Study is good for Diagnosis:
CrossSectional analytic study

What Kind of Study is good for Harm:
Cohort Study, population based case control

What Kind of Study is good for Prognosis:
Cohort study

What Kind of Study is good for Treatment:
RCT, systematic review

