Chapter 2 Methods
Research Methods
- More likely to see a FRQ on this topic (research methods) than on any other.
- Psychologists conduct research in order to solve practical problems.
o I.e. find out which method of teaching children is better.
o This type of research is known as applied research as it has a clear, practical application.
- Other psychologists conduct basic research
o Explores questions that are of interest to psychologists but are not intended to have immediate RL applications.
§ Incl. how people form their attitudes about others, how people define intelligence.
Terminology
- Good research is both valid and reliable
o Valid is when it measures what the researcher set out to measure; I t is accurate
o Reliable when it can be replicated, is consistent.
- Hypothesis and variables
o Hypothesis expresses relationship between two variables
o Variables are things that can vary among participants in research.
§ Religion, stress level, height are variables.
§ Where dependent variable depends on the independent variable.
- When testing a hypothesis, researchers manipulate the independent variable and measure the dependent variable.
o Hypotheses often grow out of theories, which aims to explain some phenomenon and allows researchers to generate testable hypotheses with the hope of collecting data to support the theory.
- Researchers need to name variables they will study and operationalize them
o When one operationalizes a variable, you explain how you will measure it.
o For example, w/ testing indep variable violent TV programs against dependent variable, people’s responses, when you operationalize the variables, need to consider what programs will be considered violent, what behaviors to be aggressive.
- Remember that a hypothesis cannot be proved, but can be supported/disproved.
Sampling
- Before investigating a hypothesis, one needs to decide what to study.
o The individuals which research will be conducted on are subjects or participants.
o The process by which subjects are selected is sampling.
§ In order to select a sample (group of subjects), one must first identify the population from which the sample will be selected.
o Population includes anyone that could be possibly selected in the sampe.
- Goal in selecting a sample is that it be representative of a large population.
o Better off with larger, more diverse population of 1000, and then randomly selecting a sample of 100.
o Random selection increases the likelihood that the sample represents the population.
- What is random?
o If one chose a sample by standing in front of the library and approaching people in a way I feel is random, I did not use random sampling.
§ Would have been less likely to approach people I did not know.
§ Also method mentioned would not represent schoo’s population.
· Not everybody walks past library on Wednesday morning.
o AR random sampling best done using a computer, table of random numbers, etc.
- Making a sample truly representative
o If I randomly select a sample of 100 people from population of 1000, sample will probably not reflect composition of school.
§ Example: If population has 500 males, 500 females, what are chances that the random sample will have the same 1:1 ratio?
· Not necessary to compute odds, however
§ Larger the sample, more likely it is to represent the population.
o So why not use all 1000 people?
§ Down side is time and money.
· Often for population psychologists, population is much larger than 10000.
§ AR large but not too large samples are optimum.
o If asked to design your own research on the AP exam, should specify the sample size and avoid using samples of extreme sizes, large or small.
- Can also increase representative trait of sample via stratified sampling
o Allows a researcher to ensure that the sample represents the population on some criteria.
§ If one thought that subjects of different racial groups might respond differently, I would want to make sure that I represented each race in the same proportion that it appears in the population.
Experimental Method
- Experiments can be divided into lab experiments and field experiments
o Laboratory experiments are conducted in a lab, a highly controlled environment.
§ Advantage is the control
o Field experiments are conducted out in the world.
§ Advantage is how it is more realistic.
- Students often equate research with experiments.
o HV know that many different kinds of research can be conducted, but only experiments can identify cause and effect relationships.
- Psychologists’ preferred method of research is the experiment because only through a carefully controlled experiment can one show a causal relationship
o Experiment allows the researcher to manipulate the independent variable and control for confounding variables
§ A confounding variable is any difference between the experimental and control conditions, except for the independent variable that might affect the dependent variable.
o AR in order to show that violent TV programs cause the subjects’ aggression, I need to rule out any other possible cause.
§ An experiment can achieve this goal by randomly assigning subjects to conditions and by using various methods of control to eliminate differences (confounding variables)
- Assignment is the process by which subjects are put into a group, experimental or control.
o Random assignment means that each subject has an equal chance of being placed into any group
§ Benefit is that it limits the effect of subject-relevant confounding variables
· Example: if subjects given choice of choosing to watch violent TV or not violent TV, it is highly unlikely that two groups would be of similar people.
o Perhaps violent people prefer violent TV and would therefore select the experimental group.
§ Even if one assigned people to groups based on a seemingly random criterion, still some possibility of subject-relevant confounding variables
o TF using random assignments lowers chance that subjects in two different groups differ. In other words, it controls for subject-relevant variables.
o Random assignment controls for subject-relevant confounding variables.
- Note: When talking about differences between groups, will be talking about group average.
o A single very aggressive subject will not throw off results of entire group.
o Idea behind random assignment is that in general, the groups will be equivalent
- If one wanted to make sure experimental and control groups were equivalent on some criterion (ex. gender, IQ scores, age), one could use group matching.
o If one wanted a group for gender, would first divide sample into males and females, then randomly assign half of each group to each condition.
o Group matching will not result in the same number of males and females within each group.
§ Rather, half of males and half of females would be in each group.
- Situation-relevant confounding variables can also affect an experiment.
o For subjects to be equal isn’t enough.
o Situations into which the different groups are put must also be equivalent except for the differences produced by the independent variable.
§ Example: If experimental group watches violent TV in large lecture hall, and control group watches programs in a small classroom, their situations are not the same.
· AR any difference may be due to confounding variables and not the independent variable changes.
§ Other situation-relevant variables include time of day, weather, presence of other people in the room.
o Making environments of two groups as similar as possible controls for situation-relevant confounding variables.
- Experimenter bias is a special kind of situation-relevant confounding variable.
o Where experimenter bias is the unconscious tendency for researchers to treat members of experimental and control groups differently to increase the chance of confirming their hypothesis.
§ Note that experimenter bias is not a conscious act.
§ If researchers purposely distort their data, it is called fraud, not experimenter bias.
o Can be eliminated by use of a double-blind procedure.
§ This occurs when neither the subjects nor the researcher are able to affect the outcome of the research
§ Can be accomplished by having researcher have someone who does not know of the subjects’ condition to interact with the subjects.
· Single blind occurs when only the subjects do not know to which group they have been assigned.
§ This minimizes the effect of demand characteristics as well as certain kinds of response or subject bias.
o Demand characteristics are cues about the purpose of the study
§ Subjects use such cues to try to response appropriately.
o Response or subject bias is tendency for subjects to behave in a certain way
§ Example: Circle the midpoint on a scale or pick the right-hand option more than the left one.
§ One kind of response bias is called social desirability
· Is the tendency to try to give politically correct answers.
o Double blinds eliminate both experimenter and subject bias.
- Experiments typically involve at least one experimental group and a control group.
o Experimental group gets the treatment operationalized in the independent variable.
o Control group gets none of the independent variable.
§ Serves as a basis for comparison.
§ Without this, would not know whether changes in experimental group are due to experimental treatment or simply to any treatment at all is impossible.
· In fact, selecting a group for monitoring alone affects their performance (i.e. monitoring workers while they work would increase their performance) > this is known as the Hawthorne effect.
o TF continuing w/ TV example.
§ Experimental group might be shown “Die Hard” (a violent TV) while control group would view “When Harry Met Sally” (a comedy, or at least some other type of television)
§ Many experiments involve much more complicated designs.
· EX: additional groups could view other types of films and view differing amounts of violent content.
- One important method of control is the placebo method
o Whenever subjects in the experimental group are supposed to ingest a drug, participants in a control group are given an inert but otherwise identical substance.
§ Allows the researchers to separate the physiological effects from the psychological effects of people thinking they took a drug called the placebo effect.
- Sometimes using subjects as their own control group is possible, a procedure called counterbalancing.
o Example: if I wanted to see how frustration affected performance on an IQ test, I could have my subjects engage in a task unlikely to cause frustration, test their IQ, then give them a frustrating task and test their Iqs again.
o HV this procedure creates possibility of order effects
§ Subjects may do better on IQ test simply because they had the virtue from having taken the first IQ test.
§ This problem can be eliminated via counterbalancing
· Having half the subjects do the frustrating task first and half the subjects do the not-frustrating task first and then switching.
- Remember that a control group is not the only way of controlling an experiment.
Correlational Method
- A correlation expresses a relationship between two variables w/o showing cause.
o Can be positive or negative.
§ Where a positive correlation means that presence of one thing predicts the presence of another
§ Negative correlation has presence of one thing predicting the absence of another.
o Correlation is anything between –1 and +1
§ Where +1 = perfect positive correlation
§ -1 = perfect negative correlation.
§ Number 0 denotes weakest possible correlation – no correlation.
o Remember that correlation does not imply causation.
- Sometimes testing a hypothesis with an experiment is impossible.
o Example: Perhaps one wants to test hypothesis that boys are more likely to call out in class than girls.
§ Cannot randomly assign subjects to conditions.
· Boys are boys, girls are girls.
§ Assignment of the independent variable, in this case, has been predetermined.
§ TF can never be able to isolate the cause of the calling out behavior.
· Could be biological difference, or many of social influences that act on gender, etc.
o AR would have conducted an ex post facto study
- An even more popular research design is the survey method.
o Contrast with experimental method: Original hypothesis that “watching violent television programs makes people more aggressive” cannot be tested using the survey method because only an experiment can reveal a cause-effect relationship.
o However, survey method can be used to investigate whether there is a relationship b/w two variables, watching violence on TV and aggressive behavior.
§ In the survey method, neither of the variables are manipulated, TF there is no independent or dependent variable.
o In the survey method, the researcher can no longer control for subject-relevant confounding variables.
§ Some people can watch a lot of violent TV, others a few.
o Survey method does not enable the researcher to determine which of these differences cause a difference in violent behavior
- Flaws of the survey method
o Controlling for situation-relevant confounding variables using the survey method is possible (bringing all subjects in one place in one time to complete survey), but it is rarely done.
§ One of advantages of survey method is that survey is mailed and completed by their convenience
§ HV since people fill out surveys in different places, times, taking different amounts of times, research will be plagued by confounding variables.
o AR determining what causes difference in violent behavior close to impossible.
o A survey will also have few people send it back, and having a low response rate, will have people unlikely to make up a representative sample.
Naturalistic Observation
- Sometimes research opt to observe subjects in natural habitats without interacting with them at al
o Goal is of a naturalistic observation is to get a realistic and rich picture of the subjects’ behavior.
o HV control is sacrificed.
- Be sure to not confuse naturalistic observation with field experiments
o In a naturalistic observation, the researchers do not impact the behavior of the participants at all.
o In a field experiment, the researcher has manipulated the indep and dependent variable and attempted to eliminate as many confounding variables as possible.
Case Studies
- Final research method is the case study method
o Used to get a full, detailed picture of one subject or a small group of subject.
o Example: Clinical psychologists often use case studies to present information about a person suffering from a particular disorder
- Allows the researcher to get the richest possible picture of what they are studying
- Does not allow the researcher to generalize the findings to a larger population.
Statistics
Descriptive Statistics
- Descriptive statistics, as the name suggests, describes a set of data.
o Example: If interested in researching what kinds of pets schoolmates have, can summarize data by creating a frequency distribution that tells how many students have each.
o Where frequency distributions can easily turn into line graphs called frequency polygons or bar graphs known as histograms.
§ Where y-axis always represents frequency, and whatever you are grasphing (i.e. pets) is the x-axis.
- One group of statistical measures is measures of central tendency
o They attempt to mark center of a distribution.
o Three common types are median, mean, and mode.
§ Obvious/common knowledge to compute.
§ A distribution may have more than one mode, and would therefore be bimodal.
o Mean is most commonly used to measure central tendency, but its accuracy can be distorted by extreme scores or outliers
§ When a distribution includes outliers, the median is often a better measure of central tendency.
- Unless a distribution is symmetrical, it is skewed.
o Outliers skew distributions
§ Where if a distribution has an extreme score that is very high, distribution is said to be positively skewed.
§ If an extreme score is very low, it is negatively skewed.
o Where a positively skewed distribution contains more low scores than high scores, opposite for negatively skewed distribution.
o In positively skewed distribution, mean is higher than the median as outliers have a much more dramatic effect on the mean than the median.
- Another type of descriptive statistical measures is variability.
o Measures include range, variance, and standard deviation.
o Measures of variability attempt to depict the diversity of the distribution.
§ Range is range, standard deviation is the square root of variance.
· Both standard deviation and variance relate the average distance of any score in the distribution from the mean.
· Higher the variance and standard deviation, the more spread out the distribution.
- Comparing scores between different distributions.
o In order to do so, can convert scores from different distributions into measures called z scores.
§ Z scores measure the distance of a score from the mean in units of standard deviation.
§ Scores below the mean have negative z scores, opposite is true for positive z scores.
o Example¨If Clarence scored 72 on a test w/ mean of 80 and deviation of 8, his z score is –1.
- Often in psychology will see reference to a normal curve
o Normal curve is a theoretical bell-shaped curve for which the area under the curve lying between any two z scores has been predetermined.
§ Where approximately 68% of scores in a normal distribution fall within one standard deviation, and almost 95% fall within two standard deviations.
· Almost 99% of scores fall within 3 standard deviations.
o Normal curve is symmetrical and knowing that three numbers given above will allow you to calculate the approximate percentage of scores falling between any given z scores.
§ Example: 47.5% of scores fall between z scores of 0 and +2.
- While z scores measure the distance of a score away from the mean, percentiles indicate the distance from 0.
o Someone who scored in the 90th percentile scored better than 90% of the people who took the test.
o Clear relationship between z score and percentile
§ Scoring in the 50th percentile means having a z score of 0
§ Scoring in the 98th percentile means an approximate z score of +2.
Correlations
- A correlation measures the relationship between two variables
o Can be positive or negative
o Where strength of correlation expressed by a correlation coefficient.
§ Range from –1 to +1.
§ Both are equally strong correlation.
o Correlations may be graphed using a scatter plot.
§ The closer the points come to falling on a straight line, the stronger the correlation
§ Where the line of best fit is the line drawn through a scatter plot that minimizes the distance that minimizes the distance of all points from the line.
Inferential Statistics
- Where descriptive statistics provide a way to summarize information about a sample, the purpose of inferential statistics is to determine whether or not findings can be applied to a large population from which the sample was selected.
o Primary goals in selecting a sample is for it to represent its population.
§ Where guaranteeing that a sample is representative of a population is impossible
o TF the extent to which the sample differs from the population is sampling error.
§ Example: if testing whether sugar impacts memorization, and you have 2 groups of 10 people, gave them lollipop, one with sugar, one without, then asked them to recall 15 words, and the results were average: 7.0 and 6.9, it is right to be skeptical.
o Differences between groups are likely due to sampling error and chance.
§ Where purpose of inferential statistics is to help psychologists decide whether findings can be applied to larger population.
§ Inferential statistical tests include t-tests, ANOVAs, and MANOVAs.
· All take into account magnitude of difference and size of the sample.
o Most important to know all the tests yield a p value
§ Smaller the p value, the more significant the results.
§ P value of .05 is the cutoff for statistically significant results.
o Where p value of 0.05 means there is a 5 percent that results occurred by chance > as a result, scientists often try to replicate results to gather more evidence that findings not by chance.
§ P value can never be 0.
APA Ethical Guidelines
- Ethical considerations a major part of research design.
- Established by the APA (American Psychological Association) for human and animal research.
o Where any type of research must first propose the study to the ethics board or institutional review board (IRB).
§ Where the proposals are reviewed for ethical violations and/or procedural errors.
o Board gives researchers permission to go ahead to revise their plan.
Animal Research
- Strict requirements on how animals must be treated in laboratory experiments.
- Requires that experiments involving animals must have:
o Clear scientific purpose.
§ Animals are chosen because they are the best option.
o Must be cared for in a humane way.
o Must acquire all subjects legally.
§ Purchased from accredited companies.
§ Wild animals must not be trapped in an inhumane manner.
o Must design experimental procedures that employ least amount of suffering feasible.
Human Research
- Informed consent – Participants must know they are involved in research and give their consent.
o If participants are deceived in any way about the nature of study, deception must not be so extreme as to invalidate the informed consent.
- No coercion in that participation must be voluntary
- Participants’ privacy must be protected – must have anonymity/confidentiality.
o Identities/actions not revealed by the researcher.
o Participants have anonymity when researchers do not collect any data that enables them to match a person’s responses with name
§ In interview studies, researcher cannot promise anonymity but can guarantee confidentiality, meaning that researcher will not identify source of any data.
- Participants not be placed at significant mental or physical risk
o This clause requires interpretation by review board.
o Consideration highlighted by Stanley Milgram’s obedience studies in 1970s
§ Participants thought they were causing significant harm/death to others.
- Participants must be told purpose of study and provided with ways to contact researchers about study results – have sufficient debriefing procedures.
0 comments:
Post a Comment