Statistics Flashcards
Intra class coefficient (ICC)
measures reliability between measures/items
>.75=good
.5-.75=moderate
Kappa
Measures intrarater reliability
>.75=excellent
.6-.75=good
.4-.6=fair
Positive predictive value
Proportion of positives that are actually positive
true positive/true positive+false negative
>10=large
5-10=moderate
2-5=small
<2=rarely important
Negative productive value
Proportion of negatives that are actually negative
true negative/true negative+falsenegative
Levels of Measurement
Nominal: categorical labels (red/yellow/blue)
Ordinal: Rank system without true meaning in spacing (MMT)
Interval: equal distances, but not related to a true zero (temperature, calendar years, IQ)
Ratio: equal distances, related to a true zero (weight, distance, age, time, blood pressure)
Strength of Evidence-lettering
A: Strong B: Moderate C: Weak D: Conflicting E: Theoretical F: Expert
Strength of Evidence–levels
Level 1: high quality RCT & prospective
Level 2: lesser quality RCT & prospective
Level 3: Case control or retrospective
Level 4: Case series
Level 5: Expert opinion
Prospective Study
watch for outcomes over time; Usually take a group of individuals and watch for the development of disease over time –level 4. Say exposure at 2 weeks, start following at 4 weeks over time—looking for disease development
Retrospective Study
study a cohort with a disease and look backwards to determine risk factors and/or protective factors (more criticized than prospective)-level 3–say exposed 1 year ago–look back at who was exposed/who developed disease later in time
Case control
RETROSPECTIVE, observation based on comparing two existing groups with different outcomes. used to determine factors that my contribute to a medical condition. compare those with and without the disease but are similar; similar to retrospective but usually starting with a group that has the disease and those that don’t and looking for the exposure variable
Cohort study
usually prospective; forward looking, planned in advance and carried over a period of time; longitudinal/observational
Inta rater
reliability with self
Inter rater
reliability between different raters
Type 1 error
Rejecting a correct null; meaning you accepted a false hypothesis and there was actually no meaningful difference between samples
Type 2 error
Not rejecting a false null–meaning your hypothesis was actually correct and there is a difference between samples (occurs more in small sample sizes)
Power
Probability that you will reject a false null–likelihood that the hypothesis is correct; .8 is standard (confidence interval is dependent on power)–narrower confidence interval =stronger study
P values
The smaller the value, the better chance the findings are a false null or the hypothesis is correct
Effect Size
Degree to which the null is false; the larger the effect size the greater the power or the more likely that the hypothesis is correct Trivial: .1-.2 Small effect: .2-.4 Moderate effect: .5-.7 Large effect: .7-2.0
Cross sectional Study
observational data collected at 1 point in time in entire population
Longitudinal
Series of observations over time
Inductive Reasoning
narrative reasoning–going from bottom down, specific to general
Deductive reasoning
Diagnostic reasoning–general to specific; testing hypothesis
Dialectical
flowing between inductive/deducitve thinking/reasoning; expert clinician