Experiments Prof’s Speech – Purple
Types of Questionnaires
- Interviewer-administered
o Face-to-face
o Telephone
- Self-administered
o Mail/email, internet
Sampling – important from the practical/conceptual standpoint
- Surveys are trying to get a representative sample (so that it can be generalized to the
greater population)
- A population is the entire pool from which a sample is drawn
o A population in this sense is not everyone who lives in a particular place
o Sampling unit – usually a person
Can be an organization, group, item, group of items, etc.
o Sampling frame – optimally, will be the same as the population
Key distinction
Probability and non-probability samples
- Probability samples
o There is an equal probability of each person being drawn (ex. Opinion polls)
o Should use probability samples if you want to get better results
o Random; everyone in the population has an equal chance at selection
Note: scientific random ≠ colloquial random
„Random‟ in science still has strategy, it is not haphazard, still
planful, but there is no set system in regards to how the selection
will take place
o Type: stratified samples
Needs a certain number of people from each group in a population
o Type: cluster samples
Identifying clusters in an area and drawing information from the cluster
- Non-probability samples
o Opposite of probability samples
o Not everyone has an equal possibility of being in the sample
o Non-probability samples are used because they are cheap and easy, but it is
difficult to extend findings from these samples to the greater population
o Type: convenience samples
Ask passersby in a mall (for example) to do an experiment
o Type: snowball samples
Participants are asked to „tell a friend‟ and those friends „tell a friend‟ too
Not everyone has a chance to participate in these samples
o Type: purposive sampling
Looking for people knowledgeable of a certain field
In terms of the scientific approach – we are still in the testing phase Some relations between variables are just a relation, they are not necessarily – one variable
causes the other variable (Example: relation between those who watch soap operas and eating
disorders – soap operas cannot cause eating disorders)
Problems with establishing causality in correlational research
Direction of influence problem
o Ex: class attendance and good grades – is it that those with greater attendance get
the higher grades? Or that those with higher grades come to class more?
o X Y or X Y
Third variable problem
o Another variable affects the result; there‟s another variable influencing both
variables
o Ex: relation between ice cream sales and number of drownings
There may seem to be some relation, but temperature is really affecting
both variables
o X – Y, but Z X and Z Y
Three things needed to establish causality
Temporal order must be correct
o The cause has to come before the outcome
Variables have to covary
No other variable is causing the outcome
o Experiments help in making sure that no other variable is causing the outcome
- Experiments can help by:
o Holding extraneous variables constant
When you hold them constant, you may still have some influence, but the
influence will be the same for all participants
If you can‟t hold them constant, use random assignment
Random assignment is when participants are assigned to one
condition or another
Independent variable is always presented first, and is the causal factor; the
independent variable causes the dependent variable
Extraneous vs confounding variable
Confounding variables are related to the independent variable
Distinctions
o Extraneous variables
Anything other than the independent variable that could affect the
dependent variable
o Confounding variables
Subset of extraneous variables
They are related to the independent variable
How experiments allow you to do this (see notes above)
Hold all variables constant Use random assignment
Present i.v. first
Present i.v. carefully and consistently, measure the d.v. rigorously
Basic experimental design(s)
One i.v. two levels
o One independent variable that has two levels, which are usually:
Experimental group
Sometimes called the treatment group
Encounter the experimental stimulus
The group where something happens
Control group
Doesn‟t receive the experimental stimulus
But is the same as the experimental group in all other ways
In drug experiments, this group will get placebos
Sometimes the two groups are just two groups that you are contrasting
with each other
o With the same time between each stage:
Experimental group administered IV measure DV
Control group measure DV
One d.v.
o One IV, one DV
Pretest- posttest design
Make sure groups are actually equivalent
Identify people with high/low pre-existing characteristics
In case of mortality
Measure change in each individual
Pretest measure – ensure that there are no pre-existing differences between the groups
involved, and are done before anything is done/administered to the groups
o Pretest measures are extra check to make sure that the groups are equivalent
o Pret
More
Less
