PSC 41 Midterm #2 Study Guide

Types of Observations

1. Observing Action (naturalistic: natural setting or systematic: research purpose) 2. Observing Performance: performance on test or tasks 3. Observing Archives: documented information

Threats to Construct Validity of Observations

1. Observer Expectancy 2. Observers can affect what they see - self-fulfilling prophecy 3. Observer Effects

Observer Expectancy

You see what you expect to see. Solutions: make observations objective & accurate: blind observer, establish inter-rater reliability, check list, time sampling

Observers can affect what they see/self fulfilling prophecy

Solution: blind observer: either unaware of hypothesis of study or which group each participant is in

Observer Effects

Being observed influences behavior so participants will act differently when they know they're being observed. Solution: be unobtrusive (hide) and allow time to acclimate

Purpose of Survey

-Attitudes, beliefs, intentions, memories, opinions

Types of Survey Questions

Open & Closed

Open Ended

Information in their own words, more open to interpretation

Closed

Forces a choice between limited options, easy to compare, bad if you dont know all/most of the possible answers

Potential Problems W/ Closed Questions

1. Make sure the categories don't overlap

2. Make sure your categories don't skip groups

3. Make sure your categories account for all/most of the possibilities (if you include "other"), you dont want most people to choose it

Types of Closed Questions

1. Yes/No 2. MC 3. Forced Alternative 4. Likert Type Scale

Forced Alternative

-Give two extremes and force a choice between them

-Give opnion by picking best out of the options

-Multiple items that tap into the same construct

-Sum the scores

Likert Scale

Graduated range of response choices w/ equal intervals

Odd number of choices

Make strong statements

Semantic Differential Format

-Rate target object using a numeric scale anchored w/ adjectives

Easy 1 2 3 4 5 Hard

Mistakes to Avoid when Writing Survey Questions

1. Avoid leading questions-word neutrally

2. Avoid double barreled items (two questions in one)

3. Avoid social desirability bias

4. Avoid long questions

5. Avoid negation (double negatives)

6. Avoid big words

7. Avoid words that may be misinterpreted

8. Question order can also affect how they respond

Best way to know your survey is effective is through pretesting

Response Sets

Mental shortcuts that influence the way people answer

Type of Response Sets

1. Acquiescence (Yay-saying): always saying yes: use reverse wording

2. Nay-Saying

3. Social Desirability

4. Faking Good/Faking Bad

5. Extremity (only extremes)

6. Fence-Sitting

Controlling For Response Sets

1. Reverse wording and coding

2. Interleave questions participants are likely to answer yes to w/ questions they're likely to answer no to

3. Look for inconsistencies

Validated Inventory

Estimate a behavior through self report by:

1. Carefully observer participants behavior and classify them into groups

2. Ask them a battery of questions

3. Determine which questions are answered differently by the different groups

4. Use those questions to classify future participants without having to observe them

Limitations to Self Report Data

-Surveys are good for opinions and attitudes, not good for behaviors

-People need to be aware of the information

-Be able to access the information (people forget)

-If you don't have to use a survey, don't

-We cant determine causality because we're often unaware of why we do things

If you have to use survey to measure behaviors

-make sure they're aware of the info (specific behaviors)

- have access to the information (recent behaviors)

External Validity

-Generalizability

-Do the results apply to the: entire population, other people, other settings, other situations

-Research w/ representative sample has a strong external validity

Error

Extent to which sample differs from the population

Types of Error

Standard Error: Random error, no two samples will be the same, there will be variation Standard Bias: Constant error, occurs if sample is selected systematically

Sample Bias

-Caused by non-random selection process/procedure: selection based on ease or any other systematic difference, self-selection

-Results in a sample that isn't representative of the population

-Reduces external validity

-Can't draw conclusions about the population

Dealing With Sample Error

1. Increase sample size: larger population = closer to your population (though not much gain beyond 500 participants)

2 Estimate using statistics-standard error

Types of Samples

Probability: representative samples v similar to population, no sample bias, strong external validity

Non-Probability: Not representative, sample not similar to population, sample bias & weak external validity

Types of Probability Samples

1. Simple Random Sampling 2. Stratified Sampling 3. Cluster Sampling 4. Multi-Stage Sampling 5. Oversampling 6. Systematic Sampling

Simple Random Sampling

-Two simple steps: 1. Identify all members of the population 2. Every member has an equal likelihood of being selected

-Representative of the population, can draw conclusions about the entire population

Stratified Sampling

1. Identify all members of the population

2. Gather information about all members of the population

3. Identify meaningful subgroups (proportions): Count subgroups in population

4. Divide population into subgroups

5. Randomly select proportional samples from each subgroup: set sample size & calculate sample size for each group

Advantages of Stratified Sampling

-Can accomplish representativeness w/ smaller random sample

-Ensures that you will not overlook a small minority group

-Can draw conclusions about the rest of the population

Cluster Sampling

-used when "natural" but relatively heterogeneous groupings are evident in a statistical population

-the total population is divided into these groups (or clusters) and a simple random sample of the groups is selected.

-Ex) college students in California: get a list of all college students (cluster) take random sample of 5 of those colleges (clusters), and then include every student of those 5 colleges in your study.

Multi-Stage Sampling

-Expansion of cluster sampling

-Cluster of colleges in Cali

-Random sample from students within in college

Oversampling

-when proportions are too small, researchers oversample the minority and then adjust the results so the members of the oversampled group are weighed to their actual proportion in the population

Systematic Sampling

-Larger population selected according to random starting point & fixed periodic interval

-Ex) choose the number 4 and 7. Start w/ the 4th person in the room and choose every 7th person

Non Probability Sampling

Not representative, but way more realistic. Samples have low external validity and can't draw conclusions about the entire population

Types of Non-Probability Sampling

Quota, Purposive, Snowball, Haphazard/Convenience

Quota

Similar to stratified sampling as in proportions. Researcher sets proportions in order to compare groups or improve representativeness. (Ex: if women only made up 1% of population but researcher wants 10 women 10 men)

Purposive

-Select specific subsets to compare

-Systematic non-random

-not attempting to make generalization

-select group that gives the effect wanted

Snowball

-Interested in a hard to access population

-Have each member recommend additional members

Convenience

What's convenient and available, most research

Improving Representativeness of non probabilistic samples

-Use selection procedures to minimize selection bias

-Sample from a broad variety of times and places

Correlation-Correlation

-Correlation Studies & Correlation Coefficient

Correlation Studies

-Describe relationship between measured variables

-Associative claims

-Correlation does not imply causation

-Variables can be continuous or categorical

Correlation Coefficient

-Statistic used to describe the relationship between two variables

-Used when both variables are measured continuously

-Can be used in correlational research or experimental research

Describing a Relationship

1. If both variables are measured continuously - look @ the relative position of each participant on two continuous variables

2. Calculate correlation coefficient - number that quantifies relationship (r)

Correlation Coefficient (r)

-r is between -1 and 1

-Indicates nature of relationship (positive r > 0 negative r < 0, no relationship r = 0)

-Strength of relationship: absolute value of r: how close the statistics are together

Relationship between variables

Positive, Negative, Curvilinear, No Relationship

Positive Linear Relationship

Increase in one variable relates to an increase in another

Negative Linear Relationship

Increase in one variable related to a decrease in another

Curvilinear Relationship

Increase in one variable relative to both increases and decreases in aother

No relationship

Y = 0 at average

Interpreting a Correlation

1. If two variables are correlated we can conclude: 1. Perhaps X & Y are related (x related to Y and Y related to X)

2. Perhaps both X and Y are related to some other variable & only related to each other in a spurious relationship

3. Perhaps it is a coincidence

-Correlation does not imply causation, impossible to tell if changes in x caused changes in Y unless you use experiments!!

Spurious Relatonship

X and Y are related to a third variable, but not directly related to each other

Evaluating Correlation Coefficient

Is this a real relationship? Statistical significance, when do we know r doesn't equal 0

How big is this relationship: effect size, r^2

Is it Significant?

-Tells you the probability that result is due to chance

Low probability that due to chance = high probability that its real

Rule for rejecting and failing to reject the null hypothesis

- p < 0.05, if p < 0.05 we reject the null hypothesis! Willing to accept 5% that we are seeing a relationship when one does not exist

Effect Size

How big is the relationship, absolute value of r

r < 0.1 = none 0.1 < r < 0.3 weak 0.3 < r < 0.5 moderate 0.5 < r < 1 strong

R^2

-Coefficient of Determination- proportion of variance shared by two variables

A ________ is the entire set of people in which the researchers are interested.

A. representative sample

B. quota sample

C. biased sample

D. population

D. Population

Essential Qualities of An Experiment

-Cause/Effect Criteria

-Definition of An Experiment

-Construct, Statistical, Internal, and External Validity

-Random Assignment

-Ruling out everything else

-Types of Experiments

-Threats to Internal Validity

-True & Quasi Experiments

Cause/Effect Criteria

-Co-occurance: