13 terms

Age-Period-Cohort effect

1. period

a. regardless of age/cohort, change in rate in particular period

2. cohort

a. regardless of age/period, shift in peak of rate according to cohort

3. age

a. regardless of cohort/period, shift in peak of rate according to age

a. regardless of age/cohort, change in rate in particular period

2. cohort

a. regardless of age/period, shift in peak of rate according to cohort

3. age

a. regardless of cohort/period, shift in peak of rate according to age

Survival analysis

1. actuarial

a. interval based on regular periods

b. censored handled by contributing 50%

2. survival

a. intervals based on event occurences

b. censored events in contribute to period they werent censored, ignored/removed from susceptible persons moving forward

a. interval based on regular periods

b. censored handled by contributing 50%

2. survival

a. intervals based on event occurences

b. censored events in contribute to period they werent censored, ignored/removed from susceptible persons moving forward

point prevalence

1. ALL cases / total pop

Cumulative incidence

1. total # events occurred during period of interest / susceptible population

2. can be same in treatment arms

2. can be same in treatment arms

Incidence rate

1. # events occurred / person time of susceptible persons

2. can be different in treatment arms

2. can be different in treatment arms

CC

1. select based on outcome then look for E

2. when OR approx. RR

a. rare disease assumption

b. CCoh b/c rep. general pop

2. when OR approx. RR

a. rare disease assumption

b. CCoh b/c rep. general pop

Research question

PI/ECO

P = who subject, what problem being addressed

I/E = what is intervention/E

C = Comparison group

O = outcome of interest

P = who subject, what problem being addressed

I/E = what is intervention/E

C = Comparison group

O = outcome of interest

Causality

1. marginal probability = prob event occurring

a. Pr (E) = a+b/N

2. Joint Prob = prob two events occuring (E and O)

a. Pr(E+O) = A / N

3. conditional prob = prov one event give another ahs occurred

a. Pr(O l E) = A / (A+B)

4. Interchangeability

a. when conditional prob not same as marginal prob, events NOT INDEPENDENT

b. when marg. prob. independent, groups interchangeable

a. Pr (E) = a+b/N

2. Joint Prob = prob two events occuring (E and O)

a. Pr(E+O) = A / N

3. conditional prob = prov one event give another ahs occurred

a. Pr(O l E) = A / (A+B)

4. Interchangeability

a. when conditional prob not same as marginal prob, events NOT INDEPENDENT

b. when marg. prob. independent, groups interchangeable

drawing up DAGs and causal pathways

1. include direction of arrows

2. pathways

a. E ->O

B. E <-AGE->O

C. E<-DEPR<-AGE-> O

D. E <-GEND.->AGE->O

E. E<-DEPR<-GEND->AGE->O

F. E<-GEND->DEPR<-AGE->O

a = main causal pathway

b to F = non causal pathways

b to E = unblocked (open)

F = blocked

control for Age

2. pathways

a. E ->O

B. E <-AGE->O

C. E<-DEPR<-AGE-> O

D. E <-GEND.->AGE->O

E. E<-DEPR<-GEND->AGE->O

F. E<-GEND->DEPR<-AGE->O

a = main causal pathway

b to F = non causal pathways

b to E = unblocked (open)

F = blocked

control for Age

measurement biases

1. sens and spec interpreted as how well we avoid misclass bias in study

2. ND masclass biases towrad null

e. D misclass can go in either direction

2. ND masclass biases towrad null

e. D misclass can go in either direction

general types of measurement bias

1. selection

a. berksonian

b. immortal time/ survival

c. lead time/length

d. over-diagnosis

e. compensating

2. info

a. recall

b. interviewer

c. observer

d. respondent

3. border line

a. medical surveillance

b. incidence/prevalence

c. temporal

a. berksonian

b. immortal time/ survival

c. lead time/length

d. over-diagnosis

e. compensating

2. info

a. recall

b. interviewer

c. observer

d. respondent

3. border line

a. medical surveillance

b. incidence/prevalence

c. temporal

Confouding

1. factor differs b/t E and NE that explains some to all of measure of effect identified

2. degree of distortion can be measrued by exces risk b/t crude and adjusted measures

3. OR, IR, rate difference: adjusted might no average out to crude as they do for RR

a. non-collapsibility, crude not weighted averages of adjusted

2. degree of distortion can be measrued by exces risk b/t crude and adjusted measures

3. OR, IR, rate difference: adjusted might no average out to crude as they do for RR

a. non-collapsibility, crude not weighted averages of adjusted

Exam tips

1. state assumptions

2. describe in as many ways as possible (additive, quantitative, EM)

3. anticipate shortcomings and remedies or explanations for decisions (sampling strategies and why chose to recruit from a particular sample)

2. describe in as many ways as possible (additive, quantitative, EM)

3. anticipate shortcomings and remedies or explanations for decisions (sampling strategies and why chose to recruit from a particular sample)