Lectures
Lecture 1: Summary last block 2
Lecture 2: Experiment 3
Lecture 3: Regression analysis 5
Lecture 4: Regression inference 8
Lecture 5: Regression assumptions and model assessment 9
Lecture 6: Regression analysis: Standardized coefficients and Prediction 12
Lecture 7: Prediction 15
Final Lecture: Recap and Q&A Assignment 3 19
1
,Lecture 1: Summary last block
Cronbach’s Alpha = internal consistency. It looks at the reliability between multiple variables.
Should be >0.70. R will give you the input on which variable you need to remove to be above
the threshold of 0.70. If you get a result of 1, it means that the answers of all the questions
are the same. If you need to remove an item, you also don’t take it into account in calculating
your means, etc.
From now on, note the statistical alternative hypothesis as Ha instead of H1 (can cause
confusion with theoretical hypothesis H1, H2, etc).
General rule for statistical interpretation: If not noted otherwise, risk level a = 0.05.
P < 0.05 = reject H0
P > 0.05 = fail to reject H0
General rule for theoretical interpretation:
P < 0,05 = confirm/validate/corroborate the researcher’s hypothesis
P > 0.05 = do not confirm
In research reports, researchers only mention whether the theoretical hypothesis is
confirmed or not. In reports you write: Results from statistical analyses showed X, therefore
H1 is validated.
Always look at direction (positive/negative, such as your r value or the means) and
significance (p-value)!
Correlation coefficient (Pearson’s r), sometimes noted with the greek letter that looks like
p, but is called ro. Between 2 interval variables, such as ease of use positively influences
technology adoption. Look at the r and p-value. r > 0 and P < 0.05 = confirm hypothesis.
T-Test compares average means between different groups
Look at the means of your groups and if it correlates with your hypothesis, and P-value
needs to be <0.05.
ANOVA compares average means between different groups (nominal >2 categories).
With the F-test you calculate the significance (P-value). If p <0.05, then at least 2 groups
show significant difference in the means. With Tukey you can calculate in which groups.
Look at the P-value in the Tukey and if there’s <0.05 then there is a significant correlation.
2
, Lecture 2: Experiment
Causality:
Reverse causality: X appears to cause Y, but it is actually Y causing Y
Third variable Z: both X and Y are caused by Z
3 conditions for causality:
1. relationship between X and Y: they vary together
2. time order: X cannot happen after Y
3. elimination of other possible causal factors: other factors need to be held constant or
controlled
Experimental design:
IV is always a nominal variable (different experimental conditions which the researcher
manipulates, he creates different versions)
- between-subject: 1 participant is assigned to 1 experimental condition (A or B)
- within-subject: 1 participant is assigned to several conditions (A and B)
You measure the DV, for example willingness to try something. Often an interval variable
(likert scale).
Context:
- lab
- online survey: in Qualtrics you can randomize participants, so some people see A,
some see B
- field work
Control extraneous factors:
- all things but the IV are the same
- participants need to be randomly assigned to groups
- measurement of other variables (such as age) for statistical control
Moderating variables changes the magnitude of the effect.
Mediation variable is more into why an effect happens.
The moderator can be measured (likert scale in survey to measure nutritional involvement)
and manipulated (in your experiment to change for example package colour).
Experiment validity:
Internal validity: conclusions about the effects of IVs on DVs are valid. -> lab studies.
How to ensure it? correct implementation of principles
- randomization, control of extraneous factors, etc.
Randomized Control Trial (RCT). Often used in medicine, for example a medicine and
placebo.
3
Lecture 1: Summary last block 2
Lecture 2: Experiment 3
Lecture 3: Regression analysis 5
Lecture 4: Regression inference 8
Lecture 5: Regression assumptions and model assessment 9
Lecture 6: Regression analysis: Standardized coefficients and Prediction 12
Lecture 7: Prediction 15
Final Lecture: Recap and Q&A Assignment 3 19
1
,Lecture 1: Summary last block
Cronbach’s Alpha = internal consistency. It looks at the reliability between multiple variables.
Should be >0.70. R will give you the input on which variable you need to remove to be above
the threshold of 0.70. If you get a result of 1, it means that the answers of all the questions
are the same. If you need to remove an item, you also don’t take it into account in calculating
your means, etc.
From now on, note the statistical alternative hypothesis as Ha instead of H1 (can cause
confusion with theoretical hypothesis H1, H2, etc).
General rule for statistical interpretation: If not noted otherwise, risk level a = 0.05.
P < 0.05 = reject H0
P > 0.05 = fail to reject H0
General rule for theoretical interpretation:
P < 0,05 = confirm/validate/corroborate the researcher’s hypothesis
P > 0.05 = do not confirm
In research reports, researchers only mention whether the theoretical hypothesis is
confirmed or not. In reports you write: Results from statistical analyses showed X, therefore
H1 is validated.
Always look at direction (positive/negative, such as your r value or the means) and
significance (p-value)!
Correlation coefficient (Pearson’s r), sometimes noted with the greek letter that looks like
p, but is called ro. Between 2 interval variables, such as ease of use positively influences
technology adoption. Look at the r and p-value. r > 0 and P < 0.05 = confirm hypothesis.
T-Test compares average means between different groups
Look at the means of your groups and if it correlates with your hypothesis, and P-value
needs to be <0.05.
ANOVA compares average means between different groups (nominal >2 categories).
With the F-test you calculate the significance (P-value). If p <0.05, then at least 2 groups
show significant difference in the means. With Tukey you can calculate in which groups.
Look at the P-value in the Tukey and if there’s <0.05 then there is a significant correlation.
2
, Lecture 2: Experiment
Causality:
Reverse causality: X appears to cause Y, but it is actually Y causing Y
Third variable Z: both X and Y are caused by Z
3 conditions for causality:
1. relationship between X and Y: they vary together
2. time order: X cannot happen after Y
3. elimination of other possible causal factors: other factors need to be held constant or
controlled
Experimental design:
IV is always a nominal variable (different experimental conditions which the researcher
manipulates, he creates different versions)
- between-subject: 1 participant is assigned to 1 experimental condition (A or B)
- within-subject: 1 participant is assigned to several conditions (A and B)
You measure the DV, for example willingness to try something. Often an interval variable
(likert scale).
Context:
- lab
- online survey: in Qualtrics you can randomize participants, so some people see A,
some see B
- field work
Control extraneous factors:
- all things but the IV are the same
- participants need to be randomly assigned to groups
- measurement of other variables (such as age) for statistical control
Moderating variables changes the magnitude of the effect.
Mediation variable is more into why an effect happens.
The moderator can be measured (likert scale in survey to measure nutritional involvement)
and manipulated (in your experiment to change for example package colour).
Experiment validity:
Internal validity: conclusions about the effects of IVs on DVs are valid. -> lab studies.
How to ensure it? correct implementation of principles
- randomization, control of extraneous factors, etc.
Randomized Control Trial (RCT). Often used in medicine, for example a medicine and
placebo.
3
