Weeks 1 And 2 Flashcards
1
Q
Pearson correlation
A
Linear regression for Gaussian data
-null is that the population correlation is 0 (no linear relationship)
2
Q
Spearman correlation
A
Linear ranking method used with extreme values and applied to regression for a more Gaussian look
3
Q
Regression equation
A
Y hat= bx+a
4
Q
When there are Equal SD for regression, blank = blank
A
B=r
5
Q
Multivariate
A
Techniques using multiple factors to remove confounding (except analysis of variance)
6
Q
Multiple linear regression
A
For predicting a numeric dependent variable with multiple variables
7
Q
Logistic regression
A
Used for 2 levels of dependent variables (yes/no, success/failure, etc)
8
Q
Proportional hazards
A
Dependent variable is the time until a certain event
9
Q
Relative risk
A
How many times more likely it is for 1 outcome vs another
10
Q
Odds ratio
A
An approx of relative risk
Used in logistic regression
11
Q
Hazard ration
A
Approx of relative risk, used in proportional hazards
12
Q
Mean duration of survival
A
Best if all subjects die; mean amount of time they live
13
Q
Median duration of survival
A
How long pts live, works better with censored data
14
Q
Case fatality rate
A
% of deaths from condition
15
Q
5 yr mortality rate
A
Proportion of deaths in a 5 yr period
16
Q
Mortality rate per person yrs of observation
A
of deaths/total pt years (alive and dead)
17
Q
Survival curves
A
Kaplan Meier, life table
18
Q
Censored events
A
If event of interest doesn’t occur by the end of the study or there is a competing cause of death, etc
19
Q
True experiment
A
Randomized design
20
Q
Crossover trial
A
Each subjects gets 2 or more tx (each subject is his own control)
21
Q
Equivalence trial
A
Shows that 2 tx or equivalent or close enough
22
Q
Non inferiority trial
A
1 tx is not worse than an existing tx
23
Q
Quasi experimental design
A
Strong element of control BUT no random assignment of individuals to groups
24
Q
Single subject multiple baseline
A
Quasi exp
-many observations, intervention, many observations
25
Group multiple baseline
Quasi exp
| -multiple baseline, intervention to group, multiple observations
26
Community trial
- quasi
- unit is the community but unit of analysis is the individual
- need verifiably comparable community
- randomly assigned intervention
27
Observational design
No intervention, only observation
| -researcher does NOT randomly assign or control for various conditions
28
Case report
Details of 1 interesting case
29
Case series
Collection of case reports to show a pattern
30
Ecological/correlational
Look at a group and try to draw causal relationship
| Ex: try to draw assoc between rate of homicide/suicide in London and the mean mental illness score in the boroughs
31
Cross sectional design/prevalence study
Exposure and disease status measure at a GIVEN time
Ex: ask if pt uses condoms CURRENTLY and see if they have an STD
Con: don't know if exposure (use of condom) was before contraction of disease or started using after contracting disease
32
Case control
Group of cases with disease compared to group without disease
33
Cohort study
Group of initially healthy pts are evaluated for exposure status and then followed to see what happens
34
Attrition bias
Loss of subjects --> distortion of experiment's effects
| -at least 80% of subjects should complete trial
35
Allocation concealment
People assessing elegibility should not be able to know what group the next entering patient will be assigned to
36
Rating bias
Inaccurate responses because of beliefs, expectations
37
File drawer bias
Non significan small studies end up being filed instead of published; can also be done on purpose by drug companies
38
Patient oriented outcomes
Disability, pain, additional surgeries all are things patients would care about
39
Surrogate marker
Easily measured indicator of disease status that might not be clinically important (disease oriented outcomes)
40
Noncompliance bias
When subjects don't take assigned meds
41
Small effects
Hard to be confident about these
| Want larger effects like higher relative risk with CI not including 1 (or not 0 for mean)
42
Non-randomized studies
Main kind of study confounding is a problem in
43
Necessary cause
Must be present for the disease to occur
44
Sufficient cause
Will produce effect when present
45
Biological plausibility
Belief in cause and effect relationship is higher if there is a known biological mechanism by which exposure can reasonably alter risk of disease development
46
Biological gradient
Greater degrees of exposure result in greater risk of death
47
Temporal sequence
Exposure was present consistently before the disease
48
Reversibility
Removal of exposure reduces risk of disease in exposed population
49
Nuremberg Code
Ethical code developed after Nazi experiments
50
Tuskegee study
Study of syphilis in AA which continued even after a tx was found
51
Common Rule
Second code of ethics that formally regulates hunam subjects research
52
Vulnerable populations defined by NIH
Children, preg women, neonates, prisoners, pts without decision making capacity, suicidal persons
53
Vulnerable population not listed by NIH
Impoverished, unemployed, AA/minorities, immigrants, elderly, LGBTQIA individ, etc
54
Therapeutic misconception
Congnitive error that participation in research will provide direct benefits to the subject (ex: when physician recomments pt to clinical study, suggesting pt will benefit)
55
Meta analysis
Math. Method of combining study results to get a combined conclusion
- provides a p value and/or CI
- can increase sample size/serve as guide for N
- deal with conflicted findings
- limitation such as difficult with assignment of different weightage
56
Qualitative research
- in depth interviews, focus groups, participant observation
- semi structured and structures
- open ended questions
57
Type 1 error (alpha)
Rejecting the null hypotheses when there is, in fact, no population difference (rejecting null when null is actually true)
58
Type II error (beta; inverse to power)
You fail to reject the null when there is, in fact, a difference in the population
59
Prevalence rate
Proportion of the population @ risk with disease at a particular point in time
60
Period prevalence
Proportion of population with disease @some point during a time interval
61
Incidence rate
Proportion of initially healthy population at risk that develops disease during pd of interest (number who develop it/#(who develop it+healthy). EXCLUDE IN DENOM (and num) PEOPLE WHO ARE SICK ALREADY
62
Attack rate
Prop of specified population that develops a disease from a specific cause like an endemic
63
Odds
Number of times event occurs/number of times it does not
64
Probability to odds
P/(1-P)
65
Odds to prob
Odds/(1+odds)
66
Standard error (SE(M))
Indicates the accuracy with which a sample mean estimates the population mean (NOT individuals)
67
Z score
Number of standard deviations from the mean
68
When are t tests employed?
Simple comparison of means
69
Unpaired t tests
2 independent samples
70
SEDoM
SE difference of means
| Sqrt[(S1^2+S22^2)/n]
71
T calculation
(Mean1-mean2)/SEDoM
72
Degrees of freedom in unpaired t
2n-2
73
Relative risk
Incidence of a disease in exposed population/incidence in an unexposed population
74
Odds ratio
Odds of disease in exposed group/odds of disease in controls
75
Attributable risk percent
Percent of people with a disease and risk factor, who got it BECAUSE of their exposure to a risk factor
76
Population attributable risk percent
Percentage of all cases of a disease due to a particular risk factor
77
Absolute risk reduction
Absolute difference between rate of bad outcome in treated and untreated groups
78
Relative risk reduction
Absolute risk reduction as a percentage of risk in untreated group (ARR/Rate in untreated group)
79
Number to treat
How many you need to treat to see a benefit in 1 person
| 1/ARR
80
Lead time bias
Apparent but false increase in survival (b/c of early detection, not actual changes in survival)
81
Sensitivity
If a person has the disease, the probability that the test will give a positive result
82
Specificity
If a person doesn't have the disease, the probability of a negative result
83
Positive predictive value
If a person gets a positive result, the probability he has the disease
84
Negative predictive value
If the pt gets a negative result, the probability he doesn't have the disease
85
Gold standard
Test/info you regard as true and correct
86
Likelihood ratio
Sensitivity/(100-specificity)
87
Power
Probability that the researcher will reject H0 when it should be rejected
88
Child mortality rate
Under 5 mortality per 1000
89
Infant mortality rate
Number of deaths of infants under the age of 1 per 1000 LIVE births
90
Neonatal mortality rate
Number of deaths of infants under 28 days of age per 1000 births
91
Maternal mortality rate
Number of women who die while preg. Or within 42 days of pregnancy termination per 100,000 live births
92
YLL
Years lives lost (when person died subtracted from ave life expectancy)
93
YLD
Years lived with disability (years of healthy life lost because of disability)
94
DALY- disability adjusted life year
YLL+ YLD
95
QUALY= Quality adjusted life year
1 QUALY= 1 yr of life in perfect health
96
Demographic transition
High fertility/high mortality as country develops
-population shift with economic development --> high fertility but low mortality
Developed nations have low fertility and aging populations
