Psych class- Experimental design, types, data, ethics

Units 1, 2, 3

Testable hypothesis

can be feasibly investigated given human, time, and resource constraints

Falsifiable hypothesis

Can be proven wrong by experimentation

Independent variable

 cause/manipulation variable

Dependent variable

effect/response to manipulation

Null hypothesis

IV has no effect on DV

Alternative hypothesis

IV does have an effect of DV

Replication

doing the same study multiple times to confirm/test validity of results ,BEST DONE BY A DIFFERENT SET OF RESEARCHERS

Recommended at least two times

 Important because it:

Builds reliability of claim, reduces chance of a fluke, corrects scientific errors, discourages p-hacking and ultimately tests generalizability 

Experimental research

Manipulation of an independent variable as a CAUSE, observation of a dependent variable as an EFFECT, allows for consistent replication and may strengthen a causal claim

Descriptive research

research including naturalistic observation, survey research, case studies

Naturalistic observation

NOT concerned with establishing causation, just describes the world as is, this was famously done by Goodall with chimpanzees, Piaget with child development 

 Survey research

can collect large amounts of data in a short time

PRO: inexpensive and easy

CON: relying on self-reporting, very prone to bias with question wording

Case study

observing a single subject in-depth, examples Little Albert fear conditioning study and Phineas Gage railroad case 

PRO: allows a very detailed and intimate picture of a subject 

CON: harder to generalize to a wider population

Correlational research

 studies the degree to which two or more variables are related 

Correlation coefficient

statistic that demonstrates correlation, r, can be positive (+1.0) or negative (-1.0)

Positive correlation

 both variables move in the same direction (+1.0)

EXAMPLE: temp increases are associated with ice cream sale increases

Negative correlation

variables move in different directions (-1.0)

EXAMPLE: as time spent on video games increases, GPA decreases

Correlation does not imply causation

a strong correlation can signify MANY causal possibilities, correlations can also be spurious 

Descriptive stats

Basic summary stats, a single stat can represent an entire set of data

Use to organize data into tables, charts, simple facts

Symmetric distribution

divided into equal halves

Positive skew

more values in lower end

Negative skew

more values in higher end 

Measures of central tendency

Mean, median, mode

Standard deviation

how far a typical value is from the mean, calculated by taking the square root of the variance

Inferential statistics

To what extent can we infer that the difference between groups is RELIABLE and REPLICABLE - is the observed difference real, or did it occur by chance

REJECT NULL if real and replicable (stat. significant), FAIL TO REJECT NULL if not 

P-value

Measure of statistical significance produced by a t-test

Alpha level

If p-value is less than or equal to reject the null, if greater than fail to reject 

Lower alpha level (0.05 or less) means a more precise result

Effect size

on a 0-1 scale, difference between groups

Difference between statistical and practical significance

Is the result statistically convincing versus  how much of a difference does it make in practice 

Metascience

study of the practice of science

Meta-analysis

systematic review of all recent research in a field, more authoritative than a traditional study 

Replication crisis

inability to consistently replicate the results of previously published studies 

Reasons for replication crisis

-publication bias towards studies with demonstrated stat significant

-questionable research practices such as p-hacking or hypothesizing after results are known

-Academia devalues replication studies

P-value hacking

running a test or manipulating data until you get a significant result

Bonferroni correction:

To avoid p-value hacking, makes researchers divide alpha level by the number of tests run 

If initial alpha is 0.05, if you want to run tests to manufacture significance, your alpha is now 0.005 so it’s harder

Solutions to replication crisis

-open science practices

-methodological reforms

-cultural shift towards integrity and away from statistical significance 

Deception

telling your subjects an abbreviated, modified, or entirely different purpose for your study than is actually true to measure a certain response

Debriefing

REQUIRED after any type of deception issued, must be done as soon as possible after data collection

Can also include giving subjects access to post-study medical or psychiatric care if needed

Ethical codes

Created after major ethical failures (such as Tuskegee study)

Nuremberg mandates informed consent, subject can stop participation at any time, concrete scientific basis for a study, and an experiment must have potential to yield societally-positive results that could not have been achieved otherwise

Belmont Report mandates beneficence and nonmaleficience, justice, and respect of rights and dignity, APA adds integrity

IRBs

Institutional Review Boards, universities and other large institutions all have them 

-responsible for giving or withholding clearance to conduct an experiment

-requires experimenters to fill out long detailed paperwork and other methods of review 

IACUCs

Institutional Animal Care and Use committee- IRB for animals, must include a veterinarian