Arms Hoorcolleges en Grasple
Inhoud
Overview analyse technieken.................................................................................................................. 2
Hoorcollege 1 – Multiple linear regression ............................................................................................ 2
Validity ..................................................................................................................................................... 2
Multiple Regression Analyses .................................................................................................................. 3
Multiple regression with dummy variables ............................................................................................. 5
Hierarchical MLR ..................................................................................................................................... 7
MLR Grasple: assumptions + rules of thumbs ......................................................................................... 9
Hoorcollege 2 – Regression: Mediation & Moderation ........................................................................ 11
Moderation............................................................................................................................................ 11
Follow up testing: simple slope analysis ............................................................................................... 12
WG 2: error’s ......................................................................................................................................... 13
Mediation .............................................................................................................................................. 14
Bootstrapping ........................................................................................................................................ 16
Hoorcollege 3 - Anova & Ancova ........................................................................................................... 18
Anova..................................................................................................................................................... 18
Ancova ................................................................................................................................................... 19
Homogenity of regression slopes: covariate ......................................................................................... 22
Effect size............................................................................................................................................... 24
Follow-up testing: post hoc en pairwise comparisons .......................................................................... 25
Grasple: Anova assumtions + Levene’s test .......................................................................................... 27
Grasple: ANCOVA Vergelijken unajusted means vs adjusted means .................................................... 28
Grasple: het interperteren van Ancova ................................................................................................. 31
Hoorcollege 4 – Factorial anova en MANOVA....................................................................................... 33
Factorial anova ...................................................................................................................................... 33
Follow up test factorial anova: Simple main effects ............................................................................. 34
Follow up testing: Contrast testing ....................................................................................................... 35
Manova.................................................................................................................................................. 36
Grasple Factorial anova: main effect en interaction effect................................................................... 38
Grasple: Assumptions manova .............................................................................................................. 40
Hoorcollege 5 – Repeated Measures en Mixed design ......................................................................... 41
Repeated measures ............................................................................................................................... 41
Repeated measures: Mauchly’s test ..................................................................................................... 42
Follow-up testing: Contrast testing in RMA .......................................................................................... 44
Mixed design ......................................................................................................................................... 44
1
,Arms Hoorcolleges en Grasple
Grasple: Repeated measures Mauchly’s test ........................................................................................ 45
Overview analyse technieken
Hoorcollege 1 – Multiple linear regression
Validity
Describe a situation that Describe a situation that
Term Definition
benefits [term]. detriments [term].
Reliability (when The extent to which a When you obtain the same When you ask some of the
measuring a measurement is free from results from a group with respondents to fill in your
construct) random measurement independent questionnaire, and there is
errors. This means that the measurements at different a lot of noise nuisance due
scores are independent of times, with the same to the construction of the
time, place and measurement tool. new tram line.
environment.
Construct The extent to which the In the case of a When a questionnaire for
validity instrument questionnaire on a sensitive measuring loneliness does
(operationalization) subject, the questions are not contain items about
succeeds in measuring the formulated in such a way as social loneliness, but only
construct and thus fits the to limit the tendency to items about emotional
conceptual definition. answer socially desirable. loneliness (content-validity).
2
,Arms Hoorcolleges en Grasple
Internal validity The extent to which the You carry out an You have developed a
research method can experiment in which programme to improve
eliminate alternative everything (except the children's spatial
explanations for an dependent variable) is perception. To investigate
effect/relationship. controlled. the effect of the
programme you use a pre-
and post-test. For both
tests you use the same
puzzles.
-> The research question
cannot be answered
correctly. Do children do
the puzzles in the post-test
better because of the
programme or because
they have remembered the
puzzles (testing)?
External validity The extent to which the The sample is You want to do research
research results can be representative of the into the extent of television
generalized to other population. watching among young
populations, settings and people. You will therefore
times. take a questionnaire with
40 boys you meet in the
shopping centre.
-> This is a limited sample,
which makes it difficult to
generalise the results to the
population (young people).
Statistical The extent to which the A researcher uses a certain A researcher reports a
validity results of a statistical data analysis technique only significant difference in
analysis are accurate and after he has checked means on the dependent
well-founded. whether the corresponding variable in an experiment,
conditions/assumptions but does not present an
have been met. effect size.
Multiple Regression Analyses
4 Measurement levels
Categorical ‘maakt groepen’
- Nominal
- Ordinal
Continuous ‘geeft een getal met numerieke waarde, een gemiddelde wat iets zegt over de variable’
- Interval
- Ratio
een MLR heeft een continious outcome (Y) en continious predictors (X) nodig.
3
, Arms Hoorcolleges en Grasple
Een simpele linear regression heeft 1 dependent variable (Y) en 1 indepentent variable (X).
Regression equation: Berekenen van een ŷ score (voorspellende waarde): b0 + b1 * de score op X +
e
Een multiple linear regression heeft 1 outcome (Y) en meedere predictors (X).
Om te kijken of de predictor past bij het model kijk je naar:
- 1. R-squared: Verklarende variantie
Dit percentage (dit cijfer doe je keer 100 voor een percentage) verteld je nu hoeveel variantie van de
afhankelijke variabele kan worden verklaard door de onafhankelijke variabele. Het overige
percentage geeft aan hoeveel variantie er niet kan worden verklaard door het model.
Assumption of linearity: wanneer je een rechte lijn kan trekken door de plot (de plot vormt een
ovaal)
• Hoe dichter de puntjes bij elkaar hoe beter je correlatie: hoe hoger je R-squared.
• Hoe verder van de lineaire lijn af hoe meer residuals.
- 2. B1: Slope of the regresion line → het effect (hoeveel Y toeneemt als X met 1 toeneemt)
4
Inhoud
Overview analyse technieken.................................................................................................................. 2
Hoorcollege 1 – Multiple linear regression ............................................................................................ 2
Validity ..................................................................................................................................................... 2
Multiple Regression Analyses .................................................................................................................. 3
Multiple regression with dummy variables ............................................................................................. 5
Hierarchical MLR ..................................................................................................................................... 7
MLR Grasple: assumptions + rules of thumbs ......................................................................................... 9
Hoorcollege 2 – Regression: Mediation & Moderation ........................................................................ 11
Moderation............................................................................................................................................ 11
Follow up testing: simple slope analysis ............................................................................................... 12
WG 2: error’s ......................................................................................................................................... 13
Mediation .............................................................................................................................................. 14
Bootstrapping ........................................................................................................................................ 16
Hoorcollege 3 - Anova & Ancova ........................................................................................................... 18
Anova..................................................................................................................................................... 18
Ancova ................................................................................................................................................... 19
Homogenity of regression slopes: covariate ......................................................................................... 22
Effect size............................................................................................................................................... 24
Follow-up testing: post hoc en pairwise comparisons .......................................................................... 25
Grasple: Anova assumtions + Levene’s test .......................................................................................... 27
Grasple: ANCOVA Vergelijken unajusted means vs adjusted means .................................................... 28
Grasple: het interperteren van Ancova ................................................................................................. 31
Hoorcollege 4 – Factorial anova en MANOVA....................................................................................... 33
Factorial anova ...................................................................................................................................... 33
Follow up test factorial anova: Simple main effects ............................................................................. 34
Follow up testing: Contrast testing ....................................................................................................... 35
Manova.................................................................................................................................................. 36
Grasple Factorial anova: main effect en interaction effect................................................................... 38
Grasple: Assumptions manova .............................................................................................................. 40
Hoorcollege 5 – Repeated Measures en Mixed design ......................................................................... 41
Repeated measures ............................................................................................................................... 41
Repeated measures: Mauchly’s test ..................................................................................................... 42
Follow-up testing: Contrast testing in RMA .......................................................................................... 44
Mixed design ......................................................................................................................................... 44
1
,Arms Hoorcolleges en Grasple
Grasple: Repeated measures Mauchly’s test ........................................................................................ 45
Overview analyse technieken
Hoorcollege 1 – Multiple linear regression
Validity
Describe a situation that Describe a situation that
Term Definition
benefits [term]. detriments [term].
Reliability (when The extent to which a When you obtain the same When you ask some of the
measuring a measurement is free from results from a group with respondents to fill in your
construct) random measurement independent questionnaire, and there is
errors. This means that the measurements at different a lot of noise nuisance due
scores are independent of times, with the same to the construction of the
time, place and measurement tool. new tram line.
environment.
Construct The extent to which the In the case of a When a questionnaire for
validity instrument questionnaire on a sensitive measuring loneliness does
(operationalization) subject, the questions are not contain items about
succeeds in measuring the formulated in such a way as social loneliness, but only
construct and thus fits the to limit the tendency to items about emotional
conceptual definition. answer socially desirable. loneliness (content-validity).
2
,Arms Hoorcolleges en Grasple
Internal validity The extent to which the You carry out an You have developed a
research method can experiment in which programme to improve
eliminate alternative everything (except the children's spatial
explanations for an dependent variable) is perception. To investigate
effect/relationship. controlled. the effect of the
programme you use a pre-
and post-test. For both
tests you use the same
puzzles.
-> The research question
cannot be answered
correctly. Do children do
the puzzles in the post-test
better because of the
programme or because
they have remembered the
puzzles (testing)?
External validity The extent to which the The sample is You want to do research
research results can be representative of the into the extent of television
generalized to other population. watching among young
populations, settings and people. You will therefore
times. take a questionnaire with
40 boys you meet in the
shopping centre.
-> This is a limited sample,
which makes it difficult to
generalise the results to the
population (young people).
Statistical The extent to which the A researcher uses a certain A researcher reports a
validity results of a statistical data analysis technique only significant difference in
analysis are accurate and after he has checked means on the dependent
well-founded. whether the corresponding variable in an experiment,
conditions/assumptions but does not present an
have been met. effect size.
Multiple Regression Analyses
4 Measurement levels
Categorical ‘maakt groepen’
- Nominal
- Ordinal
Continuous ‘geeft een getal met numerieke waarde, een gemiddelde wat iets zegt over de variable’
- Interval
- Ratio
een MLR heeft een continious outcome (Y) en continious predictors (X) nodig.
3
, Arms Hoorcolleges en Grasple
Een simpele linear regression heeft 1 dependent variable (Y) en 1 indepentent variable (X).
Regression equation: Berekenen van een ŷ score (voorspellende waarde): b0 + b1 * de score op X +
e
Een multiple linear regression heeft 1 outcome (Y) en meedere predictors (X).
Om te kijken of de predictor past bij het model kijk je naar:
- 1. R-squared: Verklarende variantie
Dit percentage (dit cijfer doe je keer 100 voor een percentage) verteld je nu hoeveel variantie van de
afhankelijke variabele kan worden verklaard door de onafhankelijke variabele. Het overige
percentage geeft aan hoeveel variantie er niet kan worden verklaard door het model.
Assumption of linearity: wanneer je een rechte lijn kan trekken door de plot (de plot vormt een
ovaal)
• Hoe dichter de puntjes bij elkaar hoe beter je correlatie: hoe hoger je R-squared.
• Hoe verder van de lineaire lijn af hoe meer residuals.
- 2. B1: Slope of the regresion line → het effect (hoeveel Y toeneemt als X met 1 toeneemt)
4
