termen hoofdstuk1 onderzoeksmethoden

empirisme

gebasseerd op data van

• direct zintuigelijke waarnemingen

• meetinstrumenten (bv vragenlijst)

onderzoek op een systematische, rigoureuze en reproduceerbare manier te doen

theorie data cyclus

top down approach

cupboard vs contact comfort

onderzoek naar hechtings gedrag

kenmerken van een goede theorie

• gesupord door data

• falsifieerbaar

• Parsimonious (“Occam’s razor”)

voorbeelden niet falsifieerbaare theorie

“De ongelovige Thomas heeft een punt”

Facilitated communication threatment believers

mertons scientific norms

• universalisme

• communality

• disintrestedness

• organized skepticism

publiceren stappen

1 je schrijft je studio op in manuscript

2 insturen manuscript naar wetenschappelijk tijdshrift

3 editor van het tijdschrift leest het

4 peer review

5 uitkomst

3 uitkomsten van publiceren

1 reject

2 revise

3 accept

werkwoorden die te maken hebben met verbanden

gaat samen met, heeft hoog risico op, correleert met, voorspelt, heeft een verband met

werkwoorden die te maken hebben met causale verbanden

veroorzaakt, vergroot, verkleint, leidt tot, verandert

ositive correlation between two variables

(Physical exercise and mood).

What are the causal models that could explain this positive correlation?

Directionality problem

Feedback relations

Confounding

Selection bias

Balans tussen externe en interne validiteit

Experimenteel onderzoek: doorgaans sterke interne validiteit maar zwakke externe validiteit

Correlationeel onderzoek: doorgaans zwakke interne validiteit maar sterke(re) externe validiteit

Balans tussen externe en constructvaliditeit

Kortere meetinstrumenten om de last voor deelnemers te verminderen en de externe validiteit te verbeteren

Kortere meetinstrumenten hebben doorgaans een zwakkere constructvaliditeit (bijv. lagere betrouwbaarheidsschattingen)

belmont report

1. Respect for persons

► Informed consent

► Protect vulnerable groups

2. Beneficence

► Risks and benefits for participants

► Risks and benefits for society

► Protecting participants’ personal information

3. Justice

► Fair balance between people participating in study and people benefitting from study

Reasons for fraude:

• Personality of researcher

• Rational choice by researcher

• Social context

Solutions for fraude:

• Regulation

• Norms

• Code of conduct

• Training

• Mentoring

experience sampling methods voordelen

• High ecological validity

• Less sensitive to memory bias and common method bias

• Ideal to study change in variables over time and within-person processes

experience sampling methods nadelen

• High participant burden

• Testing effects

bij goede vragen schrijven opletten voor

• Leading questions

• Double-barreled questions

• Double negations

• Order of questions

response baises

• Acquiescence

• Fence sitting

• Social desirability

• Faking bad (malingering)

meyhoden om low quality data te ontdekken

• Self-report items (“Did you answer the previous questions truthfully?”)

• Bogus items (“Are you born on the 30th of February?”)

• Instructed items (“Please indicate Strongly disagree for item 3”)

• Too fast respons times (less than 2 seconds / item)

• Longstring (choosing the same answer option for 9 items in a row)

• Individual respons variability (standard deviation of a respondent’s answers)

• Psychometric synonyms

• Personal reliability

• Mahalonobis D

relyable and valid observations soorten

-Unobtrusive observations

• Wait it out

• Measure the behavior’s results

• Codebook

• Interrater reliability

• Blind and double-blind design

increase respons rate

• Alert participants in advance

• Personalized surveys

• Clear goal and usefulness

• Clear source of origin

• Incentive (double-edged sword)

statistische validiteit bivariaat correlationeel onderzoek

1. What is the effect size?

2. How accurate is the effect estimate?

3. Is the effect statistically significant?

4. Has the effect been replicated?

5. Do outliers influence the effect estimate?

6. Is there range restriction?

7. Is the relationship linear or not?

voordelen within subjects design

1. Participants in all conditions are equivalent,

and thus form their own comparison group

(no risk of selection effects)

2. More power to detect significant differences

between conditions during statistical

analyses

3. Less participants needed compared to

between-subject designs

threats internal validity

• Order effects:

- Practice effect

- Carry-over effect

• Solution: counterbalancing

nadelen within subjects design

• Order effects

• Not always possible or practically feasible

- Example: study in which we test the benefits of mindfulness training

• Risk of demand characteristics because participant experiences all conditions

oplossingen voor testing

• Remove pretest

• Different instruments for pre- and posttest

• Control group

instrumentation oplossingen

• Remove pretest

• Pilot study to demonstrate equivalence of tests

• Counterbalancing of test order

oplossingen observer bais and dema,ding characteristics

• Double-blind design

• Masked design (single-blind design)

null effects oorzaken small diferences between groups

• Weak or unsuccesful manipulation of

independent variable

• Insensitive measures of dependent variable

• Ceiling and floor effects

null effects oorzaken large diferences within groups

• Measurement error

• Individual differences

• Situational variability/noise

power depends on

• Sample size (larger sample = more power)

• Effect size (larger effect size = more power)

• Alpha (smaller alpha = less power)

• Noise (more variance = less power)

problemen met interne validiteit quasi experimenten

• Selection effects

• Design confounds

• Maturation threats

• History threats

• Regression to the mean

• Attrition threat

• Testing en instrumentation threats

• Observer bias, demand

characteristics, en placebo effect

redenen om een quasi experiment uit te voeren

• Real-world opportunities

• Realism

• Ethics

• Construct validity

verschil tussen quasi experimenten en correlationeel design

• Sample is intentionally chosen in quasi-experiments

• Goal of quasi-experiments is to make causal claims

gelijkenissen tussen quasi experimenten en correlationeel design

• No random assignment

• No manipulated variables

open science practices

1. Comprehensive citation of literature, data, materials, and methods

2. Sharing data

3. Sharing the code required to reproduce analyses

4. Sharing new research materials

5. Sharing details of the design and analysis

6. Pre-registration of studies before data collection

7. Pre-registration of the analysis plan prior to analysis

8. Replication of published results

stappen meta-analyse

1. Formulate research question

2. Search for “primary studies”

3. Coding selected “primary studies”

4. Meta-analyse effect sizes of primary studies

5. Interpret meta-analytical effect size

meta-analyse procedure

• Search for primary studies:

• Search in databases + contacting colleagues

• 35 studies with 54 independent samples found

• Coding primary studies:

• Noting effect sizes (eg. r or d)

• Meta-analysis:

• Estimate “average” (weighted) effect size

strenghts meta-analyse

• Summary of scientific evidence

• High degree of credibility

weaknesses meta-analyse

• Quality of meta-analysis depends on quality of primary studies

• File drawer problem