PSYC 379 Exam 2 Review

representative sample

How well does the sample represent/ is generalizable to the broader population?

generalizability

How well can we apply the results to people outside of the lab setting?

• Depends on the nature of the sample, and if it’s representative

• Also known as external validity

time sampling

Observing at specific times

• Systematic: First day of the week

• Random: 4 randomly selected 30 min. periods

behavior sampling

Seek samples by observing behavior during “events”

• Birthday parties, PTA meetings, etc.

• Can be infrequent (natural disasters) or long (conferences).

• Provides real world context for observation.

count scale

How frequently something occurs within a specific unit of time

• Special type of ratio scale

• Can be analyzed similar to ratio or nominal data

reactivity

Changing your behavior because you know you’re being observed.

• Can be controlled by researcher concealing behavior, getting participants to habituate to researchers presense,

observer bias

Systematic errors in observation from an observer’s expectations.

population

complete group of objects (people, nonhuman animals, institutions) from which a sample is drawn on and to which results can be generalized.

• all students at UIUC

sampling frame

A complete list of all of the elements in a population (e.g., the electoral register)

• should be representative of the target population as a whole

• Helps operationalize terms

• Official enrollment list at the university

sample

a subset of a population of interest that is selected for study with the aim of making inferences to the population.

• Important that sample is representative of the larger population.

  • Psych students at UIUC

• Mean, median, and mode can all be used as descriptions of these statistics..

element

a member of a set, class, or group.

• Psychology students are a member of the college of LAS at UIUC

response rate bias

Individuals selected for the sample who don’t complete/return the survey

• Those who complete the survey might be different from those who don’t

non probability sampling

The opposite of random sampling. Elements do not have an equal chance of getting selected.

simple random sample

All members of a population have equal chances of being selected.

• Random selection of elements from pre-determined list.

successive independent samples design

A research method that involves observation of multiple random samples from a population over multiple time points.

internal consistency

Measure of reliability that checks that items are consistent with each other and are testing the same thing

• Cronbach’s coefficient alpha

Cronbach’s coefficient alpha

Measure of internal consistency/reliability. How closely related items of a group are.

discriminant validity

Two measures that are NOT CORRELATED AT ALL.

• Self esteem and height

moderating variable

Variables that affect direction/strength of correlation between 2 variables.

• For whom, under what conditions?

• What if a specific relationship is not observed at all times?

• What if the relationships between poverty, chaos, and psychological distress are not observed for all children?

external validity

How the results of the research/testing can be generalized beyond the sample

naturalistic observation

Researcher observes without attempts to change the situation.

• Goal: Describe behavior as it normally occurs.

ordinal scale

Rank order behaviors and events

• Who is rank 1st, 2nd, 3rd,

• Can use less than or greater than

• PROBLEMS

  • Distance between intervals may not be equal, creating unequal intervals (1st place finished in 2 seconds, 2nd in 3 hours

  • Can’t find averages.

ratio scale

Same as interval, except now there’s an absolute 0.

• Does not apply to most psych research, because zero doesn’t really apply here

• Examples: time, weight, age.

interobserver reliability

Do 2 or more observers agree in their observations?

• Impacted by definition of events/behaviors, characteristics of observers.

• For interval/ratio: correlation coefficient (for 2) or intraclass correlation (3 or more).

Demand characteristics

In an experiment, cues that may influence/bias participants behavior

• Only doing something because that’s what you think the researcher wants you to do.

biased samplles

when characteristics of the sample differ systematically from the population

• Can over/under represent segment of population

• Knowing helps contextualize the sample.

selection bias

Procedures result in one or more segment of the population being over/underrepresented.

• Exit polls of voters: non voters underrepresented.

convenience sampling

Any process for selecting a sample of individuals or cases that is neither random nor systematic but rather is governed by chance or ready availability.

• Form of nonprobability sampling

• Participants do not have equal chance

probability sampling

Sample of participants or cases is chosen from a larger group in such a way that each one has a known (or calculable) likelihood of being included

startified random

Divide population into subpopulations called strata, and then draw random samples from it.

• Everyone in the strata has an equal chance of selection

• Can be helpful for stratifying underrep. subpop.

  • However, might “oversample” harder to reach groups, making it not representative.

validity

“Truthfulness” of a measure. Does it measure what we’re trying to?

convergent validity

When the responses on one test/instrument is convergent with other measures

• Self esteem/self esteem scale

mediating variable

Variables used to EXPLAIN correlation between 2 variables with a third variable

• What third factor is responsible for this change?

• Chaos is a mediating variable between poverty and psychological distress

cleaning data

• Helps to get to know data better prior to analysis

• Check for errors, such as missing or impossible values, outliers

• Create summaries of your data using numbers, graphs, summaries

histogram

Helps summarize data which may help to identify

• outliers

• nonsense values,

• Important clues about the distribution

descriptive statistics

Ways to summarize and describe your data, like the mean, median, range.

inferential statistics

Statistics used to test hypotheses, such as if an association is significant or if two groups are different on an outcome

• Correlation, t-test, ANOVA, linear regression

• Mediation/Moderation (next time)

• Phi Coefficient (next time)

null hypothesis testing

1. Start with a null and alternative hypothesis

2. find a t-test statistic that that will connect to p-value

3. Find the p-value

Null hypothesis

Two groups will not have different means (H0)

• There is not an association between x and y

alternative hypothesis

Aligns with resarch hypothesis Two groups will have different means (Ha)

• There is an association between x and y

t test statistic

Compares two means of data

• Larger t value, smaller p value

• Smaller the chance you would have observed your results if null hypothesis was true

p value

Probability to which data supports null hypothesis

• Large t value, smaller p value

• Smaller p value means that null should be rejected in favor of alternative

• <.05, reject the null there is an association

• p>.05, keep the null, there is no association

continuous variables

Numerical variables that can take on any value within a given range. Can be measured with tools such as rulers or stopwatches

• Interval and ratio

• Example: the time it takes to run a mile

• Pearson correlation to find association

nominal scale

Used to classify behaviors, events, and characteristics into mutually exclusive categories

• Either in one category or another, not both

• Lowest level of measurement

• Use phi coefficient and chi square

multiple linear regression

Examining association between multiple independent variables

categorical variables

Qualitative

• Nominal ordinal

• Spearman rank order correlation

• Two fiind association, look at contingency table and compute phi

  • 
    

independent samples t test

Used to see if means are different between two independent groups

• Psychology and chemistry majors

dependent samples t test

Used to see if means are different between two dependent groups

• Measurement on the sample person, or closely linked participants such as couples/family.

One way ANOVA

Used To see if means are different between three or more independent groups

• young, middle aged, older adult.

contingency table

Shows the distrubution of one variable in rows and another in columns

• Used to examine the relationship between 2 categorical/nominal variables

phi coefficient

Measure of association between 2 categorical variables can range from 0 (completely independent) to 1 (perfect association)

• Like correlations, except ALWAYS POSITIVE

• Helps with testing independence theories

chi square test

Describes different pattern of association if independent or dependent

• Test use of balance observed/expected values to determine dependence (phi value)

participant observations

Observational method where observer actively participates in natural setting they are observing in.

• Can be disguised (subjects are unaware of observation) or undisguised (people know they are being observed)

• Challenges with reactivity, and impact of researcher on subjects

structured observation

Setting up a specific situation to observe behavior

• Done when it’s hard to observe behavior naturally

• Parent child interactions, selective attention gorilla test

• Problems when observers do not follow the same procedures across observations

interval scale

Distance between points on a scale is assumed to be equal

• Temperature

• Advantages: Can find mean and standard deviation

• Disadvantages: No absolute zero, so typing zero on a scale doesn’t mean anything.

mail survey

Quick, convenient, self-administrated survey method. Best for personal/embarrassing topics

• Challenges

  • Lots of people who receive won’t actually do it

  • Final sample may not be representative of pop

  • Little control over how survey is administered.

    • Costly

in person interviews

Survey method where researchers have more control over how survey is administered. Allows participants to ask questions, and interviewers to provide clarity

• Challenges: Must have very trained interviewer, or else potential room for bias/interferance

telephone interview

Survey method where participants complete brief surveys over the phone. Essentially cold calling

• Efficient, gives greater access to population

• Problems: Some people may have trust issues with answering over the phone

internet surveys

Research method. Efficient, low cost, and gives potential to recruit large samples.

• Can get very diverse/underrep. samples

• Problems

  • Selection bias: Only people who have internet

  • Are respondents bots?

  • Lack of control over research environment

increasing response rate

• Incentives:

  • Fixed: you will get a twenty dollar gift card

  • Or lottery

• Personal touch

  • Reminds people a human made this

• Interesting topic

• Responding takes minimal effort

cross sectional design

Selected sample from a population at ONE TIME POINT. Choose population of interest, do probability/convenience sampling, and then respondents complete survey.

• Snapshot look

• Helps describe population, make predictions

• Limitation: Can’t assess change over time. Need to make sure representative.

successive independent samples designs

Series of cross-sectional samples over time. Each sample comes from the same pop., but DIF. SAMPLE EVERY TIME

• Helps study population at different points in time

• Doing a study on incoming freshman each year.

• Challenges: Can’t really tell a lot about individual or non comparable samples.

longitudinal design

SAME SAMPLE AT DIFFERENT POINTS IN TIME. Good for seeing how individuals change over time/

• Problems: expensive/intensive, keeping people engaged over time

  • Attrition: dropping out of survey

Questionnaire reliability

Should be both reliable and valid. Can be improved by increasing number of good items, and having greater variability amongst individuals.

• Test retest: Getting the same results every time

• Internal consistency: Ideally over .75

Writing quesstionaires

• Write clear and specific.

• Don’t double barrel (have you eaten apples and oranges today?)

• Have conditional phrases (IF this were to happen…)

• No leading questions

• Use opposite q’s to avoid response bias