Join StudySoup for FREE

Get Full Access to
FSU - PSY 3213 - Study Guide - Midterm

Description

Reviews

Research Methods Exam 2 02/08/2017

Week 5-6 Notes

Measurement Concepts

Key properties of rulers:

∙ Rulers are divided into segments of unit length

o Centimeters

o Inches

∙ Units are consistent

∙ Units are objective

o Not always 1 inch.

Different people might line up the ruler slightly differently

∙ Carelessness

∙ Bias

Slight difference in difference rulers

Rulers increase measurement accuracy and consistency

Reliability

∙ The more similar the measurements are, the more reliable the measure is.

∙ Measurement error reduces the reliability of fish measurement. Ex: sober measurement =low variance

o Drunk measurement=light variance

Psychological measures

∙ We are measuring invisible or abstract things like intelligence ∙ We cannot trust our intuitions about whether or not we are doing a good job.

∙ Assessing validity becomes very important (construct validity) *look at concept map of measurement reliability and validity *reliability is necessary for construct validity

Test-retest reliability

∙ Get many participants and give each one the test twice ∙ Does each participant get similar score on both test?

o If yes, test is reliable

-Plotting test-retest reliability

-If the scatter is tight=similar scores

Correlation coefficient

-statistical parameter, determined mathematically We also discuss several other topics like What are the four ways in which evolution can occur?

Don't forget about the age old question of What are the different kinds of effectors?

-describes “strength” of a relationship between 2 variables, or how accurately one variable can predict another

∙ r ranges from -1 to 1

∙ 1=strongest positive relationship

∙ r=0 weakest relationship

∙ r=-1 strongest negative relationship

Internal consistency reliability

∙ Almost any test has multiple questions or items

∙ 1) Inter-item reliability: correlation between particular test items ∙ 2) Split half- reliability: correlation between average of items on first half of the test vs. average of items on the second half of the test ∙ 3) Interrater reliability

o Scientist try to measure things as objectively as possible o Sometimes the only available way to measure something is to make a subjective judgment Don't forget about the age old question of Which international legal response is in place to suppress terrorist attacks on embassies and diplomats?

We also discuss several other topics like If it makes sense, interpret the slope of regression line.

o Ex: Baby research

Reliability vs. construct validity

∙ Reliability

o Necessary for construct validity, but not sufficient ∙ Construct validity

o Is operational definition correct?

Kinds of construct validity

∙ Face validity (subjective)

o The content of the measure appears to reflect the content being measured

o Essentially intuition or common sense

∙ Content Validity (subjective)

o Similar to face validity: measures all the “things” that we known the construct to contain We also discuss several other topics like What happened to minnie foster?

o A measure of depression should attempt to measure all the symptoms of depression

∙ Criterion related validity (can be measured)

o Ability to predict a behavior

o A valid measure should predict membership in “known groups”

∙ Convergent validity (can be measured) If you want to learn more check out What are the three pillars of science & technology?

o Your measure gives the same results as another measure that is known already to be valid

∙ Discriminant validity (can be measured)

o Your measure should be different from other measures that are known to validly measure a different construct

Other measurement issues:

∙ Reactivity

∙ Kinds of measurement scales

Reactivity of measures

∙ Alters the variables that you are trying to measure Measurement scales

∙ Nominal

o No numerical or quantitative properties

o Classifies the levels of variable into categories/groups Species of animals

Country or origin

Major in school

Political party

∙ Ordinal

o Rank orders the levels of the variable

∙ Interval

o Differences between adjacent numerals are equal

o Zero does not mean nothing

∙ Ratio

o Numerical scale

Monday February 13, 2017 –Class Canceled

Wednesday February 15, 2017

Survey Research

-usually support frequency and association claims

Survey Data

∙ Can provide a wealth of data about attitudes, opinions, and potential causes of behavior on any number of topics

Anonymity-guarantee participants their names won’t be associated with their responses

Confidentiality-do not disclose a participants data in individuals form (report only aggregate results)

∙ Only report statistics data

With out anonymity and confidentiality respondents less willing to answer questions about very sensitive topics like marital issues or illegal practices

-Questions on questionnaires or surveys are called “items” Type of questionnaire items

∙ Open ended

o Introduces a topic and allows each participant to respond in his own words

o Advantages-information more complete

o Disadvantages

Participants may not understand what you are looking for

Answers may be too brief or unclear

Summarizing data difficult

∙ Restricted

o Limited number of specific response alternatives

o Advantages

Limited set of responses

Ordered sets of responses gives quantitative

information

o Disadvantages

Information not as rich

Missing alternatives

∙ Partially Open-Ended

o Similar to restricted items, but with an “other”

∙ Rating Scaled

o Participants rate the question on a likert scale

o Labeling the categories

Anchors

o Potential problem

Sometimes people have bias towards one end of the scale

o This is a kind of response set

o Sematic differential

Anchors are concepts, rather than agree/disagree or

very little/very much

o Advantage

Interval level information

o Disadvantage

Encourage some response sets

Response Sets

-is any tendency to always respond the same way (always yes or always 4)

Writing Good items

-use simple words, avoid jargon

-make questions short and easy to understand

-ask for one piece of information at a time

Like all measurement…

-you need to be concerned about the reliability and construct validity of your survey

-Reliability

∙ Test-Retest

∙ Split-half/inter-item reliability

-Assessing Validity

∙ content validity

∙ convergent/discriminant validity

∙ Criterion-based validity

Choosing participants and sampling

Population

∙ Large group including all potential participants

∙ May be defined in many ways

Sample

∙ Small subgroup of participants chosen from the population Sampling and generalization

-Goal of some research is to infer properties of the population from the properties of the sample

Generalization-degree to which the properties of the sample can be used to infer the properties of the population

-generalization is only possible with a representative sample ∙ sufficiently large

∙ sufficiently random

Random sample: a sample in which every member of the population has an equal change of being chosen

Nonrandom: a sample from a specialized population

Random sampling techniques

∙ Simple random sampling

o Randomly select a sample from the population

o Usually requires a list of population members

o Random digit dialing is variant used with telephone surveys ∙ Stratified Sampling

o Used to obtain a diverse sample

o Population is divided into demographic strata

o A random sample is drawn from each stratum

o Guarantees that a chosen set of population is represented ∙ Cluster sampling

o Used when population are very large

o The unit of sampling is a group rather than individuals o Groups are randomly sampled from the population ∙ Multistage sampling

o Variant of cluster sampling

o First, identify large clusters and randomly sample from the population

o Second, sample smaller units from randomly selected clusters Non-random sampling

∙ In reality, what we tend to do is sample only from members of a population that are east to find

Techniques

∙ Internet research

o Participants are self-selected volunteers, know how to use computers, have access to computers, internet savvy

o Two ways to demonstrate the validity of internet research Compare internet with non-internet results

Compare internet results with theoretical predictions

∙ Convenience sampling

o Take participants from the easiest possible source

∙ Purposive sampling

o Sample non-randomly from a population of interest

o Not a random sample because you are at a particular place at a particular time

∙ Quota sampling

o The non-random version of a stratified sample: we ensure that our sample has particular proportions of desired populations o Seek sub-populations of interest; ensure that they make up specific proportions of your sample

-The goal of most research: predict from the general to the specific -must rarer that the exact results of your experiment will be applied to the general population

Sample Size

-select an economic sample

-the weaker relationship between variables, the larger the sample needed to detect it

Power analysis

∙ Input parameters of your study

∙ Will tell you how many participants you need, of how likely you would be to detect an effect given a particular sample size

Week 7 Notes

Bivariate correlational research

Correlational designs support association claims

-claims about the relationship between measured variables Construct validity vs. statistical validity

Mehl et al. 2010

Measured variables:

Well-being

∙ Operational definition: subjective well being scale

Deep talk

∙ Operational definition: daily conversations sampled using EAR device

∙ Raters coded whether each recorded conversation was “deep” of substantive as opposed to idle chit-chat.

Construct validity

-does the deep talk measure actually measure deep talk? ∙ Reliability

∙ Face validity

Does the well-being scale adequately measure well being? ∙ Reliability

∙ Criterion based validity

∙ Convergent Validity

External validity

∙ Does this result generalize to other population?

t (-5 and 5) and r (-1,1) are both ways to express effect size

Statistical validity

∙ Effect size

o Another way to refer to how well one variable predicts another variable

o If one variable predicts another variable=large effect size o The better our predictions, the larger our “effect size”

Variance is a way to express how “spread out” values of a variable are Variance=sum of squares/degrees of freedom

Variance=SS/df

N=number of subjects of observations

Df: degrees of freedom,

SS: sum of squares

Explained variance- based on the SS distances from the predicted values to the mean

Unexplained variance- based on the distance from actual data to line of best fit

Explained Variance + Residual variance =total DV variance

Proportion of explained variance is equal to r^2 (R^2)

Statistical significance-allows us to determine if a relationship between variables has been observed by chance

-whenever we measure two variables, r will NEVER BE ZERO p=probability

Threats to statistical significance

Outliers in the data

∙ not representative of the population

∙ could indicate measurement error or unusual circumstances ∙ can change a significant result to not significant or vice versa Restriction of range

∙ If one of the variables has only a narrow range of scores, correlation is reduced

Curvilinear relationships

∙ Bivariate correlations only detect linear relationships

∙ Some curvilinear relationships will give an r of 0.

Wednesday February 22, 2017

Correlation and causality

One reason: the relationship could be caused by a third variable -when there is a third variable, the predictor and the DV will be statically correlated, but have no actual causal relationship -We can never rule out all possible third variables in a correlational design

-the first test to see if a third variable accounts for a relationship between variable A and variable B: the third variable must be correlated with both variable A and variable B

-The second test: variable A and variable B are NOT related when we hold the third variable constant

Holding a variable constant: look at only one value of the variable at a time

Multiple Correlational Research

Bivariate studies look at relationships between two variables

Statistically: multiple regression analysis

Theoretically: pattern and parsimony

Preconditions for concluding causation

∙ CoVariation

∙ Temporal precedence

∙ Internal validity (no third variable)

-Study on Barros, Silver and Stein 2009

Regression Analysis

∙ One DV

∙ Multiple predictors

∙ Beta

∙ P

P tells us if beta is significant or not

-the predictor is related to the DV when holding the other predictors constant -Mediating Variables-a variable that forms a link in a causal chain between two other variables

-Mediating variables explain WHY two variables are related ∙ A is related to B

∙ A affects C and C affects B

Moderating variables: a variable that affects the relationship between two other variables

Moderating variables place limitations on WHEN two variables are related

∙ A is related to B but ONLY SOMETIMES

∙ A is only related to B when C has a certain value