Goals of Scientific Research:
- Description of behavior
- Prediction of behavior
- Determination of the causes of behavior
- Explanations of behavior
Three Types of Scientific Studies:
- Controlled studies
- Correlational studies
- Descriptive studies
Controlled Studies (true experiments):
Random Assignment:
-Taking subjects and randomly assigning them to groups; this controls for extraneous variables -No bias
-No differences between groups
Independent Variable:
-The variable a researcher is interested in and manipulates Pex: the food given)
Dependent Variable:
-The variable the experimenter measures; outcome
Example:
Hypothesis: people who study for an exam while listening to music will score better than people who study in silence
Independent variable: the music-listening Dependent variable: test performance on the exam
2 Key Components of a Controlled Experiment:
- Random Assignment
- Identical Experimental Conditions
Random Sampling/Selection
-the initial large group was randomly selected
Stratified Sample College Example
-Randomly selected colleges and randomly select the same number of students from each college
-This is so you avoid being too specific – it needs to be generalizable
Correlational Studies
-Patterns of co-occurrence between two observed events
-Ex: cigarette smoking and cancer
-Correlation does not equal causation
-There may be a third factor – it is hard to control for something in the real world
-Correlational studies are better ethically than controlled studies
Descriptive Studies
-These studies just seek to describe an aspect of the world as it is
Design Flaws in Experimental Design:
Clever Hans
-A horse that could add and subtract, read German and answer simple questions by tapping his hoof
-The flaw: the horse would know when to stop and start tapping simply by getting cues from his owner Pextraneous factors)
Infants’ Perception of Musical Structure
-Infants turn their heads when they hear different music selections
-The flaw: the mothers influence their infants to change their head position when the mother also hears the music selection
-Solution: make the mothers wear headphones
Computers, Timing and Other Pitfalls
-The computers do not guarantee updates and other mechanical means – it is good science to check things yourself and measure things on your own too.
Number of Subjects:
Population
-The total group of people to which the researcher wishes to generalize findings
Homogeneous populations
-All individuals are similar and alike
-You won’t need to test too many people
Heterogeneous populations
-Individuals are very different
-You will need to test everyone in the population Types of Experimental Designs:
Between-Subjects Design Pindependent groups)
-Each subject is in one condition only
Within-Subjects Design Prepeated measures) -Each subject is tested in every condition
Advantages:
–smaller number of people required and you can test how each individual is affected by each manipulation
Disadvantages:
Demand characteristics:
-the subjects’ performance can be influenced by a desire to make a certain condition work better Carry-Over Characteristics:
-an effect that “carries over” from one experimental condition to another Order Effects:
-how a subject may be influenced by the order in which they do each condition
-i.e. the conditions are ordered in a certain way and a person may get influenced by stimuli presented to them
To Reduce Order Effects Prandom orders)
-Use “n factorial”
-Latin Square PN x N) or PN x 2N)
4 Principle Ethical Considerations in Using Human Subjects:
- Informed consent: subjects need to agree to do the experiment
- Debriefing: explain the experiment after it is over to the subject and answer questions
- Privacy and Confidentiality: keep the subject’s data and information confidential and stored
- Fraud: researchers must not copy or create false data or allow it to be published
Quantitative Analysis (statistical analysis):
Measurement Error:
-any difference between the observed value and the real or true value which leads to the skewing of results if not solved
-Between groups/conditions differences
-The more measurements you take = the less measurement error
Performance Error:
-The subjects will not perform identically every time
Significance Testing:
-Uses a “p value”: the probability that the experimental result could have arisen by chance -Determines whether a result is repeatable
Alternatives to Classical Significance Testing:
-Bayesian inferencing
-Effect sizes
-Confidence intervals: determines a given probability of the range of values within the population parameters
-Meta-analyses
-Conditional probabilities: the probability of an event given that another event has already occurred
Null Hypothesis
-predicts that the manipulation will have no effect at all
Qualitative Analysis (without significance testing):
-Research whose findings are not arrived at by statistical or other quantitative procedures.
-Graph data and see what patterns emerge
-Line graph Pcontinuous); Bar graph Pcategorical); Bivariate Scatter Plot Ptwo continuous variables and how one affects the other