Quantitative and Qualitative Research Techniques in the Social Sciences
Quantitative lecture 1
Data -> observation unit (rows) vs variable (columns)
Units for optimism and body length are non-commensurable, standardization gives
sometimes comparable units but comparison across studies gets lost
Qualitative lecture 1 1+2
Empirical-analytical approach (quantitative, few variables, nomothetic, standardized
methods, replicable, non-normative) vs empirical-interpretive approach (qualitative,
ideographic, holistic -> small samples, many variables, unstandardized, more difficult to
replicate)
Empirical cycle: 1. Observation (development, qualitative) 2. Induction 3. Deduction (testing,
quantitative) 4. Testing 5. Evaluation
Qualitative research is about forming meaning, using flexible research methods, and
providing qualitative findings -> grounded-theory approach data collecting + analysis again
and again
Reliability: will repeated measurements yield the same results
Measurement/internal validity: do we measure what we want to measure/are our
conclusions correct?
External validity: can we generalize the conclusions based on our sample to the population
Literature review: to find out what is already known + identify knowledge gap -> traditional
literature review (not replicable) + systematic literature review (criteria for selecting) +
meta-synthesis (qualitative, abstraction form for data) + meta-analysis (quantitative)
Research population: the group from which you select/sample research units and to which
you would like to generalize -> characteristics + subgroups. Small <1000 vs large >1000
Samples in qualitative research: small sample size, units are studied intensively, sequential
selection of research units, wide range of different perspectives/experiences
Probability samples: simple random sampling + stratified random sampling + cluster
sampling
Non-probability samples: convenience sampling + quota sampling + snowball sampling +
purposive sampling (selecting units according to the needs of study) -> deductive (sampling
driven by theoretical framework) + inductive/theoretical (sampling is interleaved with data-
collection & -analysis)
Research until you reach saturation -> diversity + available time and money
Quantitative lecture 2
Dependence techniques: model scores on L outcome (dependent) variables y as function of
scores on k predictor (independent) variables x. -> To predict the outcome on basis of scores
on predictors & to investigate the effect of predictors on outcomes
Simple regression: continuous & interval level of predictors -> one predictor
Model is a linear function, estimated such that variance is as small as possible
B head will be cov(x,y)/var(x) -> Best Linear Unbiased Estimator
Multiple correlation coefficient = R -> R^2 coefficient of determination
Mean square (MS) = sum of squares / df
F value = MS regression / MS residual
, Beta 1 = b1 * sx/sy = r(x,y) = coefficient you would get when regression standardized y on
standardized x
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Interdependence techniques: investigate interrelations among several variables c, no
distinction between the outcome and predictor variables.
Correlation: continuous & interval level variables -> two variables
Covariance: measures the extent to which positive/negative deviations from the mean on
variable x1, go together with positive/negative deviation from the mean on variable x2. ->
proportionality of deviations from the mean only make sense with an interval-scaled
variable -> disadvantage: value depends on units on the measurements scale + max. & min.
value of covariance also depends on measurement scale
Pearson correlation: divide covariance by product of standard deviations, measures linear
relationship, so interval-scaled variables. Does not depend on units of measurement scale.
Varies between 1 (perfect positive) and -1 (perfect negative), no relation is close to 0
Null hypothesis significance testing -> H0: r=0 vs H1: r is not 0 -> assumptions: independent
observations + variables normally distributed, sample obtained by simple random sampling
Fisher z transformation -> transforms r to a z-value that follows a standard normal
distribution N(0,1)
Also possible to create confidence interval (e.g. interval in which 95% of the correlations are
expected to fall)
Measure of relationship -> effect size r^2. For correlation 0.01 is small, 0.09 is medium and
0.25 is large -> coefficient of determination
Qualitative lecture 2
Qualitative interview: form of conversation in which one person restricts oneself to posing
questions concerning behaviours etc. to another person who provides the answers
Interview as a scientific data-collection method
Aim: getting an inside perspective + collecting information on factual knowledge
Clear role differentiation: interviewer vs interviewee
Unstructured interview: open-ended questions + perspective of interviewee + influence of
the interviewer
Structured interview: standardization + closed questions
Semi-structured interview has aspects of both
Interview process: construct interview guide -> pilot & modify -> grounded theory approach
-> report results
Interview guide: practical instruction for interviewer to get grip on the interview + for
standardization & preparation of interviews
Interview topics: address different aspects of the specific research questions
Interview questions: translate each topic into at least one interview questions that
interviewee can answer + make sure all parts of your specific research question are covered
in interview
Keywords: address different aspects of an interview question + follow-up on something
specific + give the interview more depth and detail
Good interview question = open-ended, short brief and clear, deals with only one topic at
the time, neutral and unbiased
Add some opening question that are related to the research, not threatening and deal with
matters the interviewee certainly know about and feel good about
Quantitative lecture 1
Data -> observation unit (rows) vs variable (columns)
Units for optimism and body length are non-commensurable, standardization gives
sometimes comparable units but comparison across studies gets lost
Qualitative lecture 1 1+2
Empirical-analytical approach (quantitative, few variables, nomothetic, standardized
methods, replicable, non-normative) vs empirical-interpretive approach (qualitative,
ideographic, holistic -> small samples, many variables, unstandardized, more difficult to
replicate)
Empirical cycle: 1. Observation (development, qualitative) 2. Induction 3. Deduction (testing,
quantitative) 4. Testing 5. Evaluation
Qualitative research is about forming meaning, using flexible research methods, and
providing qualitative findings -> grounded-theory approach data collecting + analysis again
and again
Reliability: will repeated measurements yield the same results
Measurement/internal validity: do we measure what we want to measure/are our
conclusions correct?
External validity: can we generalize the conclusions based on our sample to the population
Literature review: to find out what is already known + identify knowledge gap -> traditional
literature review (not replicable) + systematic literature review (criteria for selecting) +
meta-synthesis (qualitative, abstraction form for data) + meta-analysis (quantitative)
Research population: the group from which you select/sample research units and to which
you would like to generalize -> characteristics + subgroups. Small <1000 vs large >1000
Samples in qualitative research: small sample size, units are studied intensively, sequential
selection of research units, wide range of different perspectives/experiences
Probability samples: simple random sampling + stratified random sampling + cluster
sampling
Non-probability samples: convenience sampling + quota sampling + snowball sampling +
purposive sampling (selecting units according to the needs of study) -> deductive (sampling
driven by theoretical framework) + inductive/theoretical (sampling is interleaved with data-
collection & -analysis)
Research until you reach saturation -> diversity + available time and money
Quantitative lecture 2
Dependence techniques: model scores on L outcome (dependent) variables y as function of
scores on k predictor (independent) variables x. -> To predict the outcome on basis of scores
on predictors & to investigate the effect of predictors on outcomes
Simple regression: continuous & interval level of predictors -> one predictor
Model is a linear function, estimated such that variance is as small as possible
B head will be cov(x,y)/var(x) -> Best Linear Unbiased Estimator
Multiple correlation coefficient = R -> R^2 coefficient of determination
Mean square (MS) = sum of squares / df
F value = MS regression / MS residual
, Beta 1 = b1 * sx/sy = r(x,y) = coefficient you would get when regression standardized y on
standardized x
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Interdependence techniques: investigate interrelations among several variables c, no
distinction between the outcome and predictor variables.
Correlation: continuous & interval level variables -> two variables
Covariance: measures the extent to which positive/negative deviations from the mean on
variable x1, go together with positive/negative deviation from the mean on variable x2. ->
proportionality of deviations from the mean only make sense with an interval-scaled
variable -> disadvantage: value depends on units on the measurements scale + max. & min.
value of covariance also depends on measurement scale
Pearson correlation: divide covariance by product of standard deviations, measures linear
relationship, so interval-scaled variables. Does not depend on units of measurement scale.
Varies between 1 (perfect positive) and -1 (perfect negative), no relation is close to 0
Null hypothesis significance testing -> H0: r=0 vs H1: r is not 0 -> assumptions: independent
observations + variables normally distributed, sample obtained by simple random sampling
Fisher z transformation -> transforms r to a z-value that follows a standard normal
distribution N(0,1)
Also possible to create confidence interval (e.g. interval in which 95% of the correlations are
expected to fall)
Measure of relationship -> effect size r^2. For correlation 0.01 is small, 0.09 is medium and
0.25 is large -> coefficient of determination
Qualitative lecture 2
Qualitative interview: form of conversation in which one person restricts oneself to posing
questions concerning behaviours etc. to another person who provides the answers
Interview as a scientific data-collection method
Aim: getting an inside perspective + collecting information on factual knowledge
Clear role differentiation: interviewer vs interviewee
Unstructured interview: open-ended questions + perspective of interviewee + influence of
the interviewer
Structured interview: standardization + closed questions
Semi-structured interview has aspects of both
Interview process: construct interview guide -> pilot & modify -> grounded theory approach
-> report results
Interview guide: practical instruction for interviewer to get grip on the interview + for
standardization & preparation of interviews
Interview topics: address different aspects of the specific research questions
Interview questions: translate each topic into at least one interview questions that
interviewee can answer + make sure all parts of your specific research question are covered
in interview
Keywords: address different aspects of an interview question + follow-up on something
specific + give the interview more depth and detail
Good interview question = open-ended, short brief and clear, deals with only one topic at
the time, neutral and unbiased
Add some opening question that are related to the research, not threatening and deal with
matters the interviewee certainly know about and feel good about
