Experimental Design 1

What makes an experiment?

• examination of a causal relationship by testing a hypothesis

• It must control only the IV

• Must have at least 2 conditions to which subjects are randomly assigned

• The outcome must be measurable 

4 Classes of study

• experiment

• descriptive research

• correlational research

• quasi-experimental research

What is descriptive research?

• describes a variable without proving the cause

• may not use a hypothesis

what is correlational research?

• determines relationship between two variables

• recognises patterns, doesn’t prove cause

What is quasi-experimental research?

• similar to experiments but the IV is not manipulated by the experimenter and subjects do not have to be assigned to groups

• groups are naturally forming (eg. before and after smoking ban or having vs not having therapy already)

What is required of a hypothesis?

Must be measurable

what type of hypothesis is usually in papers?

alternative hypothesis

What is an experimental unit?

the entity which receives intervention or treatment (an individual or animal, individual neurone, cell culture, cell line, whole cage of animals)

Nuisance variable

unwanted factor affecting the outcome of a study and therefore increasing variability.

there are some ways to reduce or control these

confounders

variables which can influence both the IV and DV and so must be controlled for

4 types of study design

• between groups

• repeated measures (within groups)

• matched-pairs

• factorial design

what is Between groups study design

• different experimental units in each condition

between groups pros and cons

• no order effects

• needs more subjects in a group or to be matched due to increased variability

• nuisance variables can affect outcomes (differences between groups)

  • groups should be randomised to avoid this

what is Repeated measures design

same experimental units take part in each condition (eg. pre and post treatment) 

pros and cons of repeated measures

• can have order effects (counterbalance to control this)

• fewer subjects needed as less variability

• same subjects used so no risk of differing background characteristics

what is matched pairs?

each group has different subjects who have been matched in pairs based on important characteristics, and one of each pair is randomly assigned to each group

pros and cons of matched pairs

• two very similar groups without the risk of order effects

• time consuming 

• issues if one subject drops out

what is factorial design?

more than one factor is examined and has different levels (eg. intervention and sex) 

pros of factorial design

can examine more than one IV or include variables that we know may have an effect

Positive control

• sample that we know should show a response on our essay

• ensure that our outcome is correct but is not included in statistical analysis

Negative control

• sample that we know should not show a response to our assay

• ensure that our outcome is correct but is not included in statistical analysis

Standardisation

outcome measure may need to be standardised to reduce variability 

Biological replicates / experimental repeats

• repeats of an experiment on three different experimental units (eg. three rats)

• vary biologically

• each data point is included in analysis

technical replicates

• same experimental technique on one experimental unit (eg. three blood samples from one person)

• asses variability of technique

• quality control to detect any errors

• take averages to have one data point per experimental unit

  • this increases precision

Steps in designing an experiment

1. variables and how they are related

2. hypothesis

3. how to manipulate IV

4. sample groups, nuisance and confounding variable control

5. design

6. how is the DV measured (repeats, standardisation)

Internal validity

the design and analysis of an experiment can answer the question without bias. there is confidence that results are true to experimental context 

external validity 

whether findings can translate to other contexts