Experimental research design Flashcards
name three key experimental designs
Between-Subjects ; Matched Pair design; Within-subject design
Summarise Within-Subject design
One group of participants tested repeteadly (ie each participant experiences each different treatment or different variable, tested on the same outcome).
Reduces variability, but possible carry over effects.
Summarise Between-Subject design
Different groups of participants, each group experiences a different treatment or condition of the variable, and are tested on the same outcome.
Simple to implement, no carry over effect but introduces variation as all participants have individual differences
Summarise Matched pairs design
Matched pairs design is between subject: participants are grouped together according to relevant variables (age, sex..) each group then experiences a different treatment/condition and tested on the same outcome, aims at reducing variance between groups but time consuming
Name four main types of variables
Independent, Dependant, Extraneous, Confounding
Describe Dependant variable
Variable that is our outcome and which is influenced by the independent variable in experimental design
What is a third variable, unrelated to our experimental question, that might impact our results (influence our DV)
Extraneous variable, it can be individual related such as social class, personality traits, or more general such as time of the day the experiment takes places etc..
Confounding variable is
A variable that influences both the DV and IV, causing a spurious association between the two, renders the experiment worthless
The variable we manipulate and that the investigator controls is called..
Independent variable
What hypothesis are involved in experimental design ?
Experimental hypothesis, what we think will happen, and Null hypothesis on which we base the study, the null hypothesis is always that nothing will happen when we manipulate the IV
What are the two types of experimental hypothesis possible?
Uni directional (we predict precisely what we think will happen), or Bi directional (we predict there will be a difference, an effect, but no detail)
What is the logic of our Experiment?
Is the experimental data compatible with null hypothesis being true? We don’t explicitly mention the null hypo in our study, we just keep it in mind.
What is p value
The probability of our null hypothesis being true - it is always between 0 and 1 - we test wether any difference observed in the data are due to chance, or are significant.
P value numerical value is usually
Always between 0 and 1
If P= 0 : no chance at all the results were obtained by chance, our null hypo is false, we reject it
P= 1: results were for sure obtained by chance, we would fail to reject the null hypo
P is almost never 0 or 1 but somewhere in between
If our Ho is not true (ie: results diff obtained are unlikely to be due to chance), our P value should be
either lower or equal to 0.05 - We reject the Ho
If our Ho is true (ie our results diff are likely to be due to chance), our P value should be
Higher than 0.05 - we fail to reject the Ho
When does type 1 error in testing happens
When we reject the null hypothesis, when actually, our results diff were due to chance : False Positive - the lower the p value, the lower the risk of type 1 error
Name the 4 levels of data measurements (or scale)
Categorical (Nominal) - Ordinal (order based) - Interval - Ratio level (numerical with meaning)
3 examples of ratio level data
Response or reaction time; Age numbers; Grade system in score
3 examples of nominal data
eye colour, favourite brand of chocolate, male/female groups
How can you minimise nuisance variable in a study (3)
Standardization (holding as many of the variables constant between testing conditions as possible, only manipulating the iV nothing else) - Randomization for between subject design or Counterbalancing for within subject design
What is randomization
In a between subject design, we allocate subjects randomly to either group of the experiment, nothing meaningful is used to randomize
What is counter balancing
In a within subject design, we split our sample and each group experience each condition in a different order, this can reduce carry over effects
