MCRS PART 2 (NOTES)
WEEK 8 (FOR MONDAY)
MICRO LECTURES
1. INTERNAL-EXTERNAL VALIDITY } RESEARCH VALIDITY
● EXTERNAL VALIDITY : outside of your research
● population validity - do not exclude any groups within
your population / if you exclude then at least be
transparent
● ecological (environment) validity
● INTERNAL VALIDITY : concerns something within your research
- measurement validity
● the relationship we find should not be spurious
● concerns the validity of your conclusions ,
● concerns the certainty with which you can cause casual relationships
with your research
2. SAMPLING - PROBABILITY SAMPLING
● You need a sample that is a micro version of the entire
population
- simple random sample = each subject has the same chance of being
selected
possible bias :
- undercoverage (not everyone is included)
- sampling bias (not everyone is equally likely to be included)
- non-response bias
- response bias
TERMS :
● element/unit = single entity in a population , together all elements
form the population
, ● stratum = subject of elements form the population that share a
characteristic
● census = a numeration or count of all elements in a population
● sampling frame = a list of all the elements in a population that can be
individually identified ;
- can overlap with a census
- provides a way of contacting elements
SAMPLING ALTERNATIVES :(almost as good)
1. Random multistage cluster sample: if you don't have a good
sampling frame
you identify a large number of clusters -> random selection ->
selection from the random selection } sample
2. Stratified random sample:
you divide population in separate groups (strata)-> random selection
} sample
4. NON-PROBABILITY SAMPLE
● Same elements in the sampling frame either have no probability to
be selected or their probability is unknown .
- 4 CATEGORIES
● Convenience sampling (convenient elements-easily
accessible)- risk of bias is hightH
● Snowball sampling (samples contact new samples -
useful for closed community)- when the participants
refer new participants - risk of bias is high
● Purposive sampling (chosen sample based on the
judgment of the researcher) -risk of bias is high
● Quota sampling (distinguished according to
characteristics) - risk of bias is high
5. CORRELATIONAL DESIGNS(we just measure)\ no manipulation sos
-> Cross-sectional design = a cross-section population is considered at one
specific point- measured once
-> Time-series design = can refer to one person being measured at several
points in time
,-> Panel (time-series cross sectional) design
- same group of people
● longitudinal = any study that follows one or more participants over a
long period of time
● time-series = correlation : 1 person measured over time or groups
● repeated measures = experimental (persons measured over time in
all conditions)
CORRELATIONAL DESIGNS
ADDITIONAL MICRO LECTURE -FAM
- Prevent social desirability bias (validity threat)
● Indirect questioning
● Show it is ok to answer in a way that is not socially desirable
READINGS FOR MONDAY
- SURVEY :
● is a series of formatted questions posed to a sample of people
with the expectation that their responses will be returned
immediately or within a few days .
SUMMARY
● Most survey designs cannot assess causal relationships between
variables .
MONDAY’S LECTURE
SURVEYS :
● Capture public opinion at a point in time
● Use formatted rather than open-ended questions
, TYPES OF SURVEYS :
● Cross sectional -> 1 point in time
● Longitudinal -> predictions/changes over time
- trend : different people over time
- cohort : same type of people over time
- panel : same individuals over time
- cross-lagged panel : measures a dependent variable and an
independent variable at two points in time
CROSS-SECTIONAL OR LONGITUDINAL
● A cross-sectional study is the least expensive , but cannot determine
cause and effect
● Longitudinal studies are more expensive , but provide a better sense
of cause and effect
● To determine causality , you need a true experiment . However ,
panel studies with cross-lagged surveys may provide you with some
INDICATION of causality
- population = every person in the group you are studying
- sample = part of population selected for study
- census = study of entire population
- parameters = describe a population
- statistics = describe a sample
TUTORIAL FOR MONDAY
● What type of validity is generally speaking lower in survey design ?
- internal validity
● When is non-response a problem?
- cause the sample is not representative of the population
- Question for survey : shouldn't have effect but should have the
word CORRELATE , RELATE , IN RELATION, WHAT IS THE RELATION
BETWEEN THIS AND THIS , ASSOCIATION
WEEK 8 (FOR TUESDAY)
MICRO LECTURES
1. THREATS TO INTERNAL VALIDITY - PARTICIPANTS
WEEK 8 (FOR MONDAY)
MICRO LECTURES
1. INTERNAL-EXTERNAL VALIDITY } RESEARCH VALIDITY
● EXTERNAL VALIDITY : outside of your research
● population validity - do not exclude any groups within
your population / if you exclude then at least be
transparent
● ecological (environment) validity
● INTERNAL VALIDITY : concerns something within your research
- measurement validity
● the relationship we find should not be spurious
● concerns the validity of your conclusions ,
● concerns the certainty with which you can cause casual relationships
with your research
2. SAMPLING - PROBABILITY SAMPLING
● You need a sample that is a micro version of the entire
population
- simple random sample = each subject has the same chance of being
selected
possible bias :
- undercoverage (not everyone is included)
- sampling bias (not everyone is equally likely to be included)
- non-response bias
- response bias
TERMS :
● element/unit = single entity in a population , together all elements
form the population
, ● stratum = subject of elements form the population that share a
characteristic
● census = a numeration or count of all elements in a population
● sampling frame = a list of all the elements in a population that can be
individually identified ;
- can overlap with a census
- provides a way of contacting elements
SAMPLING ALTERNATIVES :(almost as good)
1. Random multistage cluster sample: if you don't have a good
sampling frame
you identify a large number of clusters -> random selection ->
selection from the random selection } sample
2. Stratified random sample:
you divide population in separate groups (strata)-> random selection
} sample
4. NON-PROBABILITY SAMPLE
● Same elements in the sampling frame either have no probability to
be selected or their probability is unknown .
- 4 CATEGORIES
● Convenience sampling (convenient elements-easily
accessible)- risk of bias is hightH
● Snowball sampling (samples contact new samples -
useful for closed community)- when the participants
refer new participants - risk of bias is high
● Purposive sampling (chosen sample based on the
judgment of the researcher) -risk of bias is high
● Quota sampling (distinguished according to
characteristics) - risk of bias is high
5. CORRELATIONAL DESIGNS(we just measure)\ no manipulation sos
-> Cross-sectional design = a cross-section population is considered at one
specific point- measured once
-> Time-series design = can refer to one person being measured at several
points in time
,-> Panel (time-series cross sectional) design
- same group of people
● longitudinal = any study that follows one or more participants over a
long period of time
● time-series = correlation : 1 person measured over time or groups
● repeated measures = experimental (persons measured over time in
all conditions)
CORRELATIONAL DESIGNS
ADDITIONAL MICRO LECTURE -FAM
- Prevent social desirability bias (validity threat)
● Indirect questioning
● Show it is ok to answer in a way that is not socially desirable
READINGS FOR MONDAY
- SURVEY :
● is a series of formatted questions posed to a sample of people
with the expectation that their responses will be returned
immediately or within a few days .
SUMMARY
● Most survey designs cannot assess causal relationships between
variables .
MONDAY’S LECTURE
SURVEYS :
● Capture public opinion at a point in time
● Use formatted rather than open-ended questions
, TYPES OF SURVEYS :
● Cross sectional -> 1 point in time
● Longitudinal -> predictions/changes over time
- trend : different people over time
- cohort : same type of people over time
- panel : same individuals over time
- cross-lagged panel : measures a dependent variable and an
independent variable at two points in time
CROSS-SECTIONAL OR LONGITUDINAL
● A cross-sectional study is the least expensive , but cannot determine
cause and effect
● Longitudinal studies are more expensive , but provide a better sense
of cause and effect
● To determine causality , you need a true experiment . However ,
panel studies with cross-lagged surveys may provide you with some
INDICATION of causality
- population = every person in the group you are studying
- sample = part of population selected for study
- census = study of entire population
- parameters = describe a population
- statistics = describe a sample
TUTORIAL FOR MONDAY
● What type of validity is generally speaking lower in survey design ?
- internal validity
● When is non-response a problem?
- cause the sample is not representative of the population
- Question for survey : shouldn't have effect but should have the
word CORRELATE , RELATE , IN RELATION, WHAT IS THE RELATION
BETWEEN THIS AND THIS , ASSOCIATION
WEEK 8 (FOR TUESDAY)
MICRO LECTURES
1. THREATS TO INTERNAL VALIDITY - PARTICIPANTS
