BIOST 310 Midterm Exam Study Set
The research cycle - ANSWER Research questions (about target populations)
Hypotheses (about target populations)
Data collection (creates samples)
Data analysis (about samples)
Inference (about target populations)
Identify unresolved or new questions (about target populations)
What makes a reflection good/bad - ANSWER - Distorted reflections
- blurry reflections
what makes a sample a good/bad reflection of a target population? - ANSWER - Biased
vs. representative samples
- small vs. large samples
( a good reflection in one respect can still be bad in the other)
Bias in data collection - ANSWER - bias (aka inaccuracy)
-a few routes to systematic error:
1. convenience sampling
2. voluntary response sampling
3. "found" data (ex: big data)
Example of convenience sampling - ANSWER when asked to pick a sample of graduates,
a nutrition major might inadvertently pick a disproportionate number of graduates from
health-related majors
Voluntary response sampling - ANSWER allows people to choose to be in the sample by
responding to a general invitation (ex: fan poll)
"found" data - ANSWER Data that you just found somewhere. Its authorship and
provenance are unknown. (ex: digital traces made by people that are stored by
companies and the government. also known as big data).
a sufficient condition for a representative sample - ANSWER - a sample is representative
if all members of a target population have an equal chance of being included in the
sample.
,- the simple random sample (SRS)
step 1: identify an exhaustive list of members of the target population
step 2: identify sample size (n)
step 4: randomly select n cases from the exhuastive list
step 5: observe variables of interest for all randomly selected cases
sample size and precision - ANSWER - recall: representative samples are not functions
of sample size
- two goals of statistical inference (parameter estimation and hypothesis testing)
- two functions of sample size in statistical inference (precision in parameter estimation
and power in hypothesis tests)
-optimal sample size
case reports and case series (the design in brief) - ANSWER - case report: in-depth
description of a single subject's health (ex: health symptoms, results of clinical workups,
etc)
- case series: description of a set of case reports that seem to resemble each other
case reports and case series (strengths) - ANSWER exploratory, useful for identifying
hypotheses about new health conditions
case reports and case series (limitations) - ANSWER not question-driven, not collected
in a systematic, representative way
cross-sectional surveys (the design in brief) - ANSWER - collect a sample from the target
population at a single point in time.
- should be represenative (random draw) but not "randomized"
- for each subject, observe value of health variable
- for each subject, observe value of risk factors and other variables of interest
- compare prevalence of disease between groups according to risk factor divisions
cross-sectional surveys (strengths) - ANSWER - useful for quantifying prevalence of risk
factors or of disease
-useful for quantifying associations between risk factors and diseases
-large, inexpensive samples
- straightforward data collection, mgmt
, -fewer ethical constraints
cross-sectional surveys (limitations) - ANSWER -specific to a single point in time:
generalization ill-advised
-cant assess temporal relationships --> association, not causation
- may miss unobserved (common-cause or lurking) variables --> association, not
causation
- nonresponse (selection) bias
(prospective) cohort studies (design in brief) - ANSWER -recruirt/enroll a sample of
individuals
- should be representative of that cohort (ex: random draw)
- for each subject, observe value of risk factors and of other variables of interest
- follow subjects over time with 1+ follow-up to assess health status and possible
changes in other variables (time-varying covariates or TVCs)
- longitudinal study designs and prospective cohort study designs intersect
recruit/enroll a sample of individuals - ANSWER (a) from a cohort (a group who have
something(s) in common, such as birth cohorts/generations)
(b) who do not (yet) have a health condition of interest
(prospective) cohort studies strengths
*this also works for retrospective cohort studies* - ANSWER - great for estimating
latency period (typical duration, range of variability)
-great for estimating incidence of disease (# or proportion of new cases)
-can assess temporal relationships between risk factor and disease --> suggestive of
causation (but not conclusive bc of common cause).
(prospective) cohort studies limitations - ANSWER - may miss unobserved
The research cycle - ANSWER Research questions (about target populations)
Hypotheses (about target populations)
Data collection (creates samples)
Data analysis (about samples)
Inference (about target populations)
Identify unresolved or new questions (about target populations)
What makes a reflection good/bad - ANSWER - Distorted reflections
- blurry reflections
what makes a sample a good/bad reflection of a target population? - ANSWER - Biased
vs. representative samples
- small vs. large samples
( a good reflection in one respect can still be bad in the other)
Bias in data collection - ANSWER - bias (aka inaccuracy)
-a few routes to systematic error:
1. convenience sampling
2. voluntary response sampling
3. "found" data (ex: big data)
Example of convenience sampling - ANSWER when asked to pick a sample of graduates,
a nutrition major might inadvertently pick a disproportionate number of graduates from
health-related majors
Voluntary response sampling - ANSWER allows people to choose to be in the sample by
responding to a general invitation (ex: fan poll)
"found" data - ANSWER Data that you just found somewhere. Its authorship and
provenance are unknown. (ex: digital traces made by people that are stored by
companies and the government. also known as big data).
a sufficient condition for a representative sample - ANSWER - a sample is representative
if all members of a target population have an equal chance of being included in the
sample.
,- the simple random sample (SRS)
step 1: identify an exhaustive list of members of the target population
step 2: identify sample size (n)
step 4: randomly select n cases from the exhuastive list
step 5: observe variables of interest for all randomly selected cases
sample size and precision - ANSWER - recall: representative samples are not functions
of sample size
- two goals of statistical inference (parameter estimation and hypothesis testing)
- two functions of sample size in statistical inference (precision in parameter estimation
and power in hypothesis tests)
-optimal sample size
case reports and case series (the design in brief) - ANSWER - case report: in-depth
description of a single subject's health (ex: health symptoms, results of clinical workups,
etc)
- case series: description of a set of case reports that seem to resemble each other
case reports and case series (strengths) - ANSWER exploratory, useful for identifying
hypotheses about new health conditions
case reports and case series (limitations) - ANSWER not question-driven, not collected
in a systematic, representative way
cross-sectional surveys (the design in brief) - ANSWER - collect a sample from the target
population at a single point in time.
- should be represenative (random draw) but not "randomized"
- for each subject, observe value of health variable
- for each subject, observe value of risk factors and other variables of interest
- compare prevalence of disease between groups according to risk factor divisions
cross-sectional surveys (strengths) - ANSWER - useful for quantifying prevalence of risk
factors or of disease
-useful for quantifying associations between risk factors and diseases
-large, inexpensive samples
- straightforward data collection, mgmt
, -fewer ethical constraints
cross-sectional surveys (limitations) - ANSWER -specific to a single point in time:
generalization ill-advised
-cant assess temporal relationships --> association, not causation
- may miss unobserved (common-cause or lurking) variables --> association, not
causation
- nonresponse (selection) bias
(prospective) cohort studies (design in brief) - ANSWER -recruirt/enroll a sample of
individuals
- should be representative of that cohort (ex: random draw)
- for each subject, observe value of risk factors and of other variables of interest
- follow subjects over time with 1+ follow-up to assess health status and possible
changes in other variables (time-varying covariates or TVCs)
- longitudinal study designs and prospective cohort study designs intersect
recruit/enroll a sample of individuals - ANSWER (a) from a cohort (a group who have
something(s) in common, such as birth cohorts/generations)
(b) who do not (yet) have a health condition of interest
(prospective) cohort studies strengths
*this also works for retrospective cohort studies* - ANSWER - great for estimating
latency period (typical duration, range of variability)
-great for estimating incidence of disease (# or proportion of new cases)
-can assess temporal relationships between risk factor and disease --> suggestive of
causation (but not conclusive bc of common cause).
(prospective) cohort studies limitations - ANSWER - may miss unobserved
