Milgram Experiment

What affects compliance?

(Obedience to authority study at Yale

 Teacher punishes "learner" with shocks)

What affects compliance?

1.  authority of experimenter

2. if learner is in other room (if can see, less compl.)

3.   force hand on shock plate, less compliance

4.  If teacher could choose voltage, never over 45 v.

*punishers didn't show l.t. negative effects of their disobedience

*Obedience rates haven't changed in the 40 years  after the study

Men and women show = rates of obedience

*Keep participant identities secret

*Use subject #s not names in datat files

*Need special permission to if you plan to share audio/video

*Be careful how data is coded, stored, used

*Use care w/ sensitive study topics (sex, drugs, divorce, etc.) or when studying special populations (i.e. kids)

APA code of ethics 1

(aka: the ethical principles of psychologists and code of conduct) 5 general principles

1. beneficence
2. responsibility
3. integrity
4. justice
5. respect for rights and dignity of others

APA ethics code 2

10 ethical standards-To what do they apply?

APA Ethics code 3

Name the 10 ethical standards

1.  institutional approval
2. informed consent to research
3. informed consent for recording voice/image in research
4. client/patient, student and subordinate research participants
5. dispensing with informed consent for research participation
6. deception in research
7. debriefing
8. ethics and animal research
9. reporting results
10. plagiarism, publication credit

• purpose
• right to decline/ withdraa, even after participation has begun
• forseeable consequences of declining/ withdrawing
• risks, discomfort, adverse effects
• prospective research benefits
• limits of confidentiality
• incentives for participation
• whom to contact w/ ??? about research/rights

1. naturalistic observation in a public place w/ no anticipation that the recording will be usedin a manner that wil cause personal i.d. or harm
2. research design uses deception and consent for the use of the recording is obtained during debriefing

1. protect from adverse consequences of declining or w/drawal
2. if participating for credit, give other way to earn

where research won't cause harm/distress and is

• the study of normal classroom/ed practices, etc, in a normal class setting
•  anonymous ?aire, nat.obs., or archival res., where loss of anonymity won't cause subject liability, damage $ standing, employability , or reputation & confidentiality is protected
• study of job/org effectiveness in a normal busn. setting w/ no risk of dmg. to participants' employability and confid. is protected

• no excessive inducements that might coerce partic.
• if offering prof. services as inducemt., clarify nature of servs., risks, obligations, limitations

• Only if justified by potential value of research & if other non-deceptive measures aren't feasible
• don't deceive about expected pain or emotional distress
• reveal deception asap ( preferably @ end of their participation, no later end of data collection)

• nature, results, conclusion of research, correct any misconceptions
• if scientific/humane values justify delaying/ w/holding this info, take measures to reduce trisk of harm
• if you're aware that procedures have harmed a participant, take meas. to minimize the harm

• rules much les strict
• expected to be humane: minimize pain, discomfort, illness, infection, etc.
• can subject to pain, distress or privation only when alternative is unavailable and the goal is justified by scientific, educational or applied value
• if an animal is to be terminated, done as quickly and painlessly as poss.

misrepresentation

(reporting research results, plagiarism, publ. credit)

• fabricating data is fraud
• if you find errors after publ., make effort to correct
• Plag: misrepresenting another's work as your own

belmont report

(3 principles)

• beneficence
• justice
• autonomy and respect for persons

• refers to need for research to to maximize benefits and minimize harmful effects of participation
• risk-benefit analysis: process of calculating both the possible risks and benefits

• physical harm (rare in psych research)
• psychological stress: unfavorable feedback, asked about traumatic events, view disturbing events, view disturbing photos
• loss of confidentiality or privacy(don't observe in settings where privacy would be invaded)

eliminate/minimize risks

inform of potential risks beforehand

if potential for lasting harm exists, research must have overriding value

make sure subj. can contact exper. in case of harm, seek remedies

confidentiality/privacy

exempt from IRB review

only anonymous questionnaires, surveys, and ed tests

naturalstic observation in public places where there is no threat to anonymity

IRB at the inst. has a procedure by which the researcher can apply for exempt status.

risks of harm to participant are no greater than risks encountered in daily life or routine Phys./Psych. tests

Examples:

1.  recording routine physiological data in such a way that privacy is not violated
2. moderate exercise by healthy volunteers
3. research on individual/group behavior, or characteristics of ind., in which researcher does not manipulate participants' behavior and the research won't involve stress to participants

subject to thorough review by IRB

requires complete informed consent, full IRB

if long study, IRB must review @least 1x/yr. 

participants should receive all information that might influence their decision whether to participate

• special populations (children, special needs)

• coercion- any procedure that limits an individual's freedom to consent (big benefits or perceived punishment for refusal to participate, as with subordinate res. participants)
• Withholding info and deception (studies have shown that informed consent procedures increase perceptions of control in stress experiments and can affect conclusions. )
• 47% of studies use deception
• deception is decr. in social psych. 

addresses issues of fairness in receiving the benefits of  research and bearing the burdens of accepting risks.

(Tuskegee)

simulation of a real-world setting,

(can be used to look at conflict between people, driving behavior, jury deliberation, etc.)

Zimbardo study

experimenter describes a situation, asks participant what s/he would do in the situation or what s/e imagines others would do

Not considered a good alternative to deception (no emotional investment, subj. may behave accdg. to how s/he thinks exper wants,  people can't accurately predict their behavior or the behav. of others

a decision rule that is determined before we collect the data: the point at which we decide that our results are significant

The probability that our results could be due to error in our sample.  If it's very unlikely that our observed result was due to error, then we reject the null hypothesis.

Anova

a statistical significance test for determining whether two or more means are significantly different.

F is the ratio of systematic variance to error variance.

F=difference between groups 

variance within groups

One way anova

Repeated measures anova

one way is used for between-subjects designs (like independent samples t-test)

repeated measures is used for within subjects designs (like paired samples t-tests)

like t-test, in that both differences in group means and the variability within groups  matters

BUT, no restriction on number of groups or # of IVs

Ratio or interval scale

F(1, 38)=10.89, p<.05

Include between and within dfs

Fvalue, p-value, in italics

Planned comparison: compare only the means in which specific differences were predicted by the hypothesis, rather than testing all possibilities.

 OR

Post Hoc Comparison tests: use special significance tests designed to control for increase in Type I errors

(example: Tukey's test)

Choose the test: 

(One IV, 2 groups)

Nominal scale data

ordinal scale data

Interval or Ratio scale data

Ord: independent groups= Mann-whiney U test

      repeated measures=wilcoxon's T or the sign test

int/ratio: independent groups=t-test for

independent samples

                repeated measures=paired samples t-test

Choose the test

(one IV, 3 groups or more)

nominal scale data

ordinal scale data

interval/ratio scale data

Choose the test

 (2 or more IVs)

Nominal scale data

ordinal scale data

Interval/ratio data

nom:chi square

ord: no approp. test

int/rat: 2 way analysis of variance

a CT stat tells us what the sample as a whole (average) is like

mean:approp. only when scores are meas. on an interval/ratio

median: score that divides the data set in half (ordinal, interval, ratio)

Mode:  only appropriate scale for nominal scale; does not use the actual values, but indicates the most frequently occuring value.

If skewed, median/mode is better indication of central tendency

a number indicating the amount of spread in a distribution  of scores

standard deviation:(s/ SD) indicate the average deviation-derived by 1st calculating the variance.  Small when most people have scores close to the mean.  b/c larger as more people have scores that lie further from the mean; uses actual values: suitable only for interval/ ratio scores

variance: (s²)(the variance is the SD squared)

range: difference between the highest and lowest scores

IV levels represented on horizontal (x) axis

DV levels represented on vertical (y) axis (mean displayed here)

bar graphs for nominal data on x axis (no model, model), line graphs for numerical data on x-axis: line drawn between points represents relationship between variables

correlation coefficient

Pearson product-moment  correlation coefficient

cc: describes how strongly variables are related

ppmcc: used when both variables have interval or ratio scale properties: pearson r

values range 0 to +-1: strength and direction of relationship (0= no relat/+-1=perfect relationship)

 does not show curvilinear relationship: to check for this, make a scatterplot 

 refers to the strength of the association between the variables (pearson r cc is one indicator of effect size)

+/- not used for ES, only 0 to 1

ES is useful in that it gives us a consistent scale of values across all types of studies

.10 to .20: small ES

.30: medium

.40+: large

sometimes useful to report squared correlation (if r=.5, r²=.25)   This changes r into a percentage, representing the % of variance in one variable accounted for by the second variable.  The rest of the variation is accounted for by other variables. 

calculations used to predict a person's score on one (criterion) variable when we already know her score on another (predictor) variable

(Y=a+bX): Y is the (criterion) variable we want to predict, X is the known score (predictor variable) , a is a constant and b is a weight (slope of the line created with this formula)

used to combine a number of variables to increase the accuracy of prediction of a given outcome \ criterion variable; higher correlation vs. one varible alone

Ex: grad school applicants rated on (1) college grades, (2) GRE scores, and (3) letters of recomendation

Y=a+b₁X₁+b₂X₂+...b_nX_n

R²=percentage in variability in the criterion variable that is accounted for by the combined set of predictor variables

researchers face the third-variable problem in non-experimental research when a third variable may be responsible for the relationship between the 2 variables of interest

Part corr. is the technique that provides a way of statistically controlling third variables (influence of 3rd variable is "partialed out of" the orig. correlation, approximation of relat. btw. the two variables of interest without the 3rd variable)  Compare correlation with 3 variables to correlation w/ 2.

Structural Equation Modeling

Path analysis 

Path coefficients

SEM:  set of techniques that specify a set of causal relats. amg. variables using the non-experimental method.  

Path analysis:  structural equation modeling technique used to study the model; stats analysis provides path coeffs. which indicate the strength of a relat. on a 0 to 1 scale

used to determine whether we can infer that the diff. in the sample reflects differences in the population

 I.S. give the probability that the diff. btw. means reflects random error, rather than a real diff. 

(t test)

null hypothesis

research hypothesis

statistical significance

NH: population mean of the no-model group is equal to the population mean of the model group

RH:  pop. mean of the no-model group is not= to the pop. mean of the model group

NH is rejected when there is a very low possibility of the observed research results  occurring due to random error

SS:a significant result is one that has a low probability of occurring if the population means are equal; probability required for significance is the alpha level

sampling distribution reflects all the possible outcomes we would expect if we compare the means  of the 2 groups and the null hypothesis is correct

need to calculate the value of t under the null hypothesis.  If the t has a low probability of occurrence (.05), the the null hypoth. is rejected 

t=    group difference (M₁-M₂)

      within-group variability (random error)

when comparing 2 means, df=N₁+N₂-2

one-tailed t: hypothesis specified the predicted direction 

two-tailed t:  hypothesis did not specify a predicted direction of difference

.05 significance level is the critical point beyond which 2.5% of the positive values and 2.5% of the negative values lie

analysis of variance

F test

 an extension of the t-test- if there is a study w/ only one IV w/ 2 groups, then F and t are nearly identical, F=t² 

F-test is  a ratio of 2 types of variance: systematic and error variance

effect size

r=the square root of  [t²/(t²+df)]

systematic variance

error variance

SV: (aka between group variance) the deviation of the group means from the grand mean, or the mean score of all individuals in all groups

EV: (aka within group variance) the deviation of the individual scores in each group from their respective group means

sv/ev=F  the larger the F ratio, the more likely it is that the results are significant

one of the effect size estimates that can be used when comparing 2 means. Expresses effect size in terms of standard deviations. (A d value of 1 tells you that teh means are 1 stand. dev. apart)

d=(M₁- M₂)/ the square root of [(SD₁²+SD₂²)]

an interval of values defines the most likely range of actual population values. An interval has an associated confidence interval:  A 95% c.i. says we are 95% confident that the population parameter of interest lies within the range.  99% interval would provide greater certainty, but would be larger (less precise). 

 Size of the interval is related to the size of the sample, also (larger sample size, narrower c.i., bc sample is more likely to reflect the true mean)

1) want to be confident that your results are reliable (would get the same results in repeated samples)

2)The alpha level you choose indicates how confident you want to be in your results alpha=.05 means we're 95% sure of the reliability of your findings

3)you are most likely to obtain signif. results w/ larger sample size

4)you are most likely to obtain signif. results when the effect size is large (difference btw. groups is large, variability of scores within groups is small)

we reject the null hypothesis, but it is actually true

Occurs when we we obtain a large F by chance

αlpha=probability of a type I error

want this low, because consequences are seen as more serious than cons. of a type II error

probability of a Type I error can be changed by changing the alpha level

Occurs when we fail to reject the null hypothesis, when the research (alternative) hypothesis is true

probability of a Type II error= Βeta

we want this to be low.  Influenced by 3 factors:

1)significance level (lower s.l. increases beta-difficult to reject the null)

2) sample size-easier to detect real diffs. w/ a bigger sample

3)effect size-if e.s. is large, type II error is unlikely BUT a small effect size may not be signif. w/ a small sample size.

population

                 H₀ is true                 H₀ is false

reject H₀     Type I error             Correct decision

                    (alpha)                     (1- beta)

probability of correctly rejecting the null hypothesis

Power = 1- p (type II error)

remember, this is affected by alpha, sample size and effect size 

IV: nominal

DV: nominal

IV: nominal, 2 groups (male-female)

DV: Interval/ratio (GPA)

IV: Nominal (3 groups) (study time: low, med, high)

DV: Interval/ ratio  (test score)

IV: interval/ratio (optimism score)

DV: Interval/ratio score (sick days last year)

IV: nominal (2+ variables)

DV: (Interval/ratio)

IV:  Interval/ratio (2+ variables)

DV: Interval/ratio

1) college students-restricted population; young, intelligent, developing identity, need for approval, unstable peer relationships

2) volunteer-more educated, in need of approval, social, higher socioec. status

3)gender-need to recognize possible differential interpretation of men & women  

4) Locale-research done in specific educational/geographic settings may not generalize

detailed information about how your study was conducted (enough info. for replication)

1) overview of study

2)participants (how recruited, what incentives were given)

3)apparatus-if special equipment was used

4) procedure- instructions given to participants, how IV was manipulated, how the DV was manipulated, how you conducted the study, measurement techniques used

description of analysis (don't explain findings)

state alpha here

describe coding

state which statistical test was used, justify using that test to support a hypothesis

summarize each finding in words and with statistical phrases

in APA, placed at the end of the paper

don't repeat the data

implications of study.  summarize original purpose and expectations of the study, state whether results were consistent w/ expectations

relationship between your results and past results

Include your own criticisms of the study

suggestions for future research