Table of contents
LECTURE II.1 - Experimental Research 2
LECTURE II.2 - Comparative Research Designs 4
LECTURE II.3 - Basic statistics 5
LECTURE II.4 - Descriptive statistics I - Averages 6
LECTURE II.5 - Descriptive statistics II - What spreads tells us about data 7
LECTURE II.6 - Data transformation and normal distribution 9
xLECTURE II.7 - Sampling and statistical inference 10
LECTURE II.8 - Confidence intervals 12
LECTURE II.9 - Statistical significance 14
1
,LECTURE II.1 - Experimental Research
The anchoring effect is the effect you experience when already given information affects
your answer (although the information was useless)
The independent variable may also be referred to as the x-variable, the explanatory variable
or the cause.
The dependent variable may also be referred to as the y-variable, the effect variable or the
outcome variable.
There are basic principles of experimental research, which include:
- A systematic testing of causal claims
- A controlled intervention in which the researcher “plays God”
- There is an isolation of effects, which leads to ceteris paribus (a term that means “all
else equal”)
In experimental research the random assignment is important. This is not random recruiting,
but rather it is a random assignment of the participants into groups (experimental or control).
This may help;
- to ensure comparability
- to minimise other factors (which will eliminate variable bias)
- to prevent reactivity (when the participants do not know which group they are
assigned to => they are blind)
- to prevent the Rosenthal effect (when the researcher do not know which participants
go into which groups => he/she is (double)blind => this will prevent the researcher
from having expectations)
A post-test is a test taken after the treatment (both groups) to infer whether the outcome has
been affected treatment (one can compare the post-test of the experimental and the control
group). Thus, it is a measure of the outcome variable. It allows the answering to the following
question; Was the intervention effective? Did randomisation work well=
A pre-test is a test taken before the treatment (both groups) both to infer whether there are
confounding factors at play (if an “improvement” has occurred within the control group) and
whether the treatment group changes over time. This is of course in additions to the
inferences which the post-test allows.
If one wants a more complex research design, one may want to use the four-group-design.
This design is meant to prevent any bias the pre-test may cause. Thus, this design has four
groups (two experimental and two control groups), where only one of the control groups and
one of the experimental groups have a pre-test.
Experimental research is the most positivistic design.
2
, When conducting experimental research, the researcher should be concerned with the level
of internal validity; is the outcome actually a result of the treatment? To enhance internal
validity, there are certain things the researcher could consider:
1. Are the groups comparable?
a. They are if they are blinded + if they are randomised
2. Are the participants treated equally?
a. This is achieved (with a greater likelihood) if the researcher is blind
3. Are the effects of time possible?
a. It is best that the effects of time is excluded, which is best noticed through the
pre-test/post-test design
4. May the repetition of the test affect the outcome?
a. This can be eliminated through the Solomon Four Group Design.
To ensure internal validity one is faced with a trade-off issue. First, having repeated
measures comes with the cost of the repetition affecting the outcome. Second, to eliminate
the issue of repetition affecting the outcome/having more control groups come with the cost
of more resources needed.
Also, when conducting experimental research, the researcher should be concerned with the
level of external validity; can the inferences be generalised to the wider population? To
enhance external validity, there are certain things the researcher could consider:
1. How have the participants been samples?
a. If the research is not random, the researcher should be explicit about the
direction of the bias
2. Do participants react to the experimental situation?
a. The participants must be blinded as much as possible (however, this does
pose some ethical considerations)
3. The issue of ecological validity; does the research simulate the real-life situation?
To ensure external validity one is, also, faced with a trade-off issue. First, to recruit a
representative sample comes with the cost of more resources needed/non-available
resources. Second, to ensure the research is controlled comes at the cost of creating a
real-life context.
Some ethical considerations are important when conducting experimental research:
1. When deception in this research design often is inherent, debriefing afterwards is
essential.
2. One must be able to justify any unequal treatment.
Field or randomised controlled trials (RCTs) concerns controlled treatment in the field.
Quasi-experiments or natural experiments do not have control groups. Here there are
natural variations on the treatment (e.g. only some parts of East Germany received West
German broadcast => did this enhance the possibility of protest in these areas?). Natural
experiments can not have true randomisation - rather they have as-if randomisation through
case-selection.
3
LECTURE II.1 - Experimental Research 2
LECTURE II.2 - Comparative Research Designs 4
LECTURE II.3 - Basic statistics 5
LECTURE II.4 - Descriptive statistics I - Averages 6
LECTURE II.5 - Descriptive statistics II - What spreads tells us about data 7
LECTURE II.6 - Data transformation and normal distribution 9
xLECTURE II.7 - Sampling and statistical inference 10
LECTURE II.8 - Confidence intervals 12
LECTURE II.9 - Statistical significance 14
1
,LECTURE II.1 - Experimental Research
The anchoring effect is the effect you experience when already given information affects
your answer (although the information was useless)
The independent variable may also be referred to as the x-variable, the explanatory variable
or the cause.
The dependent variable may also be referred to as the y-variable, the effect variable or the
outcome variable.
There are basic principles of experimental research, which include:
- A systematic testing of causal claims
- A controlled intervention in which the researcher “plays God”
- There is an isolation of effects, which leads to ceteris paribus (a term that means “all
else equal”)
In experimental research the random assignment is important. This is not random recruiting,
but rather it is a random assignment of the participants into groups (experimental or control).
This may help;
- to ensure comparability
- to minimise other factors (which will eliminate variable bias)
- to prevent reactivity (when the participants do not know which group they are
assigned to => they are blind)
- to prevent the Rosenthal effect (when the researcher do not know which participants
go into which groups => he/she is (double)blind => this will prevent the researcher
from having expectations)
A post-test is a test taken after the treatment (both groups) to infer whether the outcome has
been affected treatment (one can compare the post-test of the experimental and the control
group). Thus, it is a measure of the outcome variable. It allows the answering to the following
question; Was the intervention effective? Did randomisation work well=
A pre-test is a test taken before the treatment (both groups) both to infer whether there are
confounding factors at play (if an “improvement” has occurred within the control group) and
whether the treatment group changes over time. This is of course in additions to the
inferences which the post-test allows.
If one wants a more complex research design, one may want to use the four-group-design.
This design is meant to prevent any bias the pre-test may cause. Thus, this design has four
groups (two experimental and two control groups), where only one of the control groups and
one of the experimental groups have a pre-test.
Experimental research is the most positivistic design.
2
, When conducting experimental research, the researcher should be concerned with the level
of internal validity; is the outcome actually a result of the treatment? To enhance internal
validity, there are certain things the researcher could consider:
1. Are the groups comparable?
a. They are if they are blinded + if they are randomised
2. Are the participants treated equally?
a. This is achieved (with a greater likelihood) if the researcher is blind
3. Are the effects of time possible?
a. It is best that the effects of time is excluded, which is best noticed through the
pre-test/post-test design
4. May the repetition of the test affect the outcome?
a. This can be eliminated through the Solomon Four Group Design.
To ensure internal validity one is faced with a trade-off issue. First, having repeated
measures comes with the cost of the repetition affecting the outcome. Second, to eliminate
the issue of repetition affecting the outcome/having more control groups come with the cost
of more resources needed.
Also, when conducting experimental research, the researcher should be concerned with the
level of external validity; can the inferences be generalised to the wider population? To
enhance external validity, there are certain things the researcher could consider:
1. How have the participants been samples?
a. If the research is not random, the researcher should be explicit about the
direction of the bias
2. Do participants react to the experimental situation?
a. The participants must be blinded as much as possible (however, this does
pose some ethical considerations)
3. The issue of ecological validity; does the research simulate the real-life situation?
To ensure external validity one is, also, faced with a trade-off issue. First, to recruit a
representative sample comes with the cost of more resources needed/non-available
resources. Second, to ensure the research is controlled comes at the cost of creating a
real-life context.
Some ethical considerations are important when conducting experimental research:
1. When deception in this research design often is inherent, debriefing afterwards is
essential.
2. One must be able to justify any unequal treatment.
Field or randomised controlled trials (RCTs) concerns controlled treatment in the field.
Quasi-experiments or natural experiments do not have control groups. Here there are
natural variations on the treatment (e.g. only some parts of East Germany received West
German broadcast => did this enhance the possibility of protest in these areas?). Natural
experiments can not have true randomisation - rather they have as-if randomisation through
case-selection.
3
