PHCY 500 Biostats

Variable

something that can change

Independent variable

what you manipulate

Dependent variable

What you measure

Nominal variable

named categories

ordinal variable

ordered categories

Interval variable

ordered, equal intervals, no true 0

Ratio variable

ordered, equal intervals, true zero

Continuous

any value (ex: height)

Dichotomous

two options (ex: yes/no)

Discrete

countable (ex: # ER visits)

Population

entire group of interest

Sample

subset of the population

descriptive statistics

summarize data

inferential statistics

draw conclusions about a population from a sample

mean

average

median

middle value

mode

most frequent value

Standard deviation

spread around mean

Interquartile range

middle 50% spread

range

min-max

frequency

how often something occurs

probability

likelihood of an event

normal distribution

symmetric, bell-shaped

binomial distribution

discrete, two outcomes, abnormal distribution

95% confidence interval

range where true population parameter is expected to lie 95% of the time

null hypothesis

no effect or difference

alternative hypothesis

effect or difference exists

alpha

probability of type 1 error (commonly 0.05)

Type 1 error

Saying there is an effect when there isn’t (false positive)

Type 2 error

Saying there is no effect when there is one (false negative)

power

probability of detecting an effect is there is one

power analysis

determines sample size needed to achieve desired power

why is a sample size important

larger samples= more accurate & reliable results & higher power

Bias

systematic error that skews results

validity

measures what it was supposed to measure

reliability

consistency of results

sensitivity

ability to correctly detect positives (catch the condition)

specificity

ability to correctly detect negatives (rule out the condition)

correlated variables

variables that change together

parametric statistics

use when data is numeric & normally distributed

Nonparametric statistics

use when data is not normal or categorical/nominal

one-sample t-test definition

nonparametric equivalent

compare sample mean to a known value

Wilcoxon signed-rank test

two-sample t-test definition

nonparametric equivalent

compare means of 2 independent groups

Mann-Whitney U test

paired t-test definition

nonparametric equivalent

compare means of 2 related groups (ex: before/after)

Wilcoxon signed-rank test

one-way ANOVA definition

nonparametric equivalent

compare 3 of more groups

Kruskal-Wallis test

Bonferroni correction

adjusts alpha when making multiple comparisons to reduce type 1 error

p < alpha

significant difference exists

p ≥ alpha

no significant difference detected

correlation

relatedness between two variables

Pearson correlation (r)

nonparametric equivalent

measures linear relationship between continuous, normally distributed variables

Spearman correlation (ρ)

r or ρ > 0

positive relationship

r or ρ < 0

negative relationship

r or ρ = 0

no relationship

linear regression

ability to predict one variable from another (independent vs dependent)

statistical controls

controls remove the influence of other factors to see the true effect

R²

proportion of variation explained by the model

contingency table

shows frequence of outcomes for 2 categorical variables

proportion

part of a group with a characteristic

odds

ratio of event happening vs not happening

odds ratio

how many times more likely (compares odds of an event between 2 groups) (ex—> odds=success/failure)

relative risk

risk of outcome (compares probability/risk of an event between 2 groups) (ex—> risk=success/total)

chi-square test

test of observed & expected frequencies

Fischer exact test

similar to chi-square test but used when expected frequencies are small (usually n<5)

McNemar test

test of paired frequencies (ex: frequency before/after)

Logistic regression

dichotomous (yes/no, adherent/not)

Multilevel model

used when data is grouped

Survival analysis

analyzes time to event (ex: death, relapse)

Proxy variable

when you can’t measure something so you measure something else as a substitute