Biostats SCOPE Objectives

case control studies

analytical observational study design where researchers identify a group of patients with and without an outcome and follow data backwards in time to determine if they were exposed to a risk factor RETROSPECTIVE

case report

descriptive observational study design that involves a clinical observation in a single patient

case series

descriptive observational study design that involves a clinical observation of a small group of patients

cohort study

analytical observational study design in which researchers identify a group of patients exposed and not exposed to a risk factor and follow data forward in time to determine if they develop an outcome may be RETROSPECTIVE or PROSPECTIVE

controlled clinical trials

prospectively and directly compares measures, and quantifies differences in outcome between an intervention and control

what is the best study method to use to determine cause and effect relationships

controlled clinical trials

cross-sectional studies

analytical observational study in which researchers identify a group of patients with and without an outcome and determine if they’ve been exposed to a risk factor at a single point in time INVOLVES CONCURRENT DATA COLLECTION

measures of central tendency

describes how the data is summarized → mean, median, and mode

mean

average used in describing continuous data; greatly affected by outliers

median

mid-most value (50th percentile) used in describing ordinal or continuous data; not affected by outliers

mode

most frequently occurring data point; also highest point of a peak on a frequency distribution; used in all types of data but is the only descriptor for nominal data

not affected by outliers

measures of variability

describes how the dependent variable is spread around the median or mean (i.e. mean ± st. deviation)

range

interval between the highest and lowest values in data group; used for ordinal and continuous data

interquartile range

directly related to the median and describes the 25th and 75th quartile; used for ordinal, interval, or ratio data

stable measure unaffected by outliers

standard deviation

measures the amount of variation or dispersion in a set of data points (sample)

variance

measures the average of the squared differences between each data point and the mean (describes how far individual points differ from the mean)

inferential stats

using the values of a sample to make an informed guess or generalizations about the population

measures of distribution

used for continuous variables (interval or ratio); shows frequency of all possible values of a variable or characteristic

central limit theorem

when equally sized samples are randomly drawn from a non-normal distribution, the plotted mean values from each sample will approximate a normal distribution as long as if the non-normal wasn’t due to outliers

parametric data

mean, median, and mode are the same point; occurs with continuous data

non-parametric data

mean, median, and mode are all different points; occurs with nominal or continuous data

list characteristics of parametric data

• bell-shaped curve

• mean = median = mode

• symmetrical

• zero kurtosis/skewness

• AUC = 1

alpha

probability of making a type I error (willing to accept 5% error threshold)

type I error

risk of finding a statistically significant difference when there is no difference between treatment groups caused by chance (false positive)

beta

probability of making type II error; the chance investigators wish to take for missing an effect between treatment and control groups

type II error

finding no difference when a true difference exists (false negative)

effect size/ delta

size of the difference or change to be detected

smallest difference considered clinically significant

p-values

the calculated chance that a type I error has occurred in a study

used to determine statistical significance

what does it mean to be statistically significant

the study demonstrates evidence against the null hypothesis; can be reasonably sure the difference between groups is not due to chance

confidence intervals

the range of numerical values which we can be certain the unknown/true population value lies; can estimate effects in populations based on data from individual samples

power

the ability of a study to detect/see a significant difference between treatment groups if one exists

(1-beta)

as power increases, …

probability of making a type II error decreases

reject the null hypothesis if…

p </= alpha

fail to reject the null hypothesis if…

p > alpha