PSY-2700: Research Ethics categories, Central Tendency, Type I and II Errors, and different tests

Consequentialist theories

“Good actions lead to good results”

Deontological ethics

Duty-based ethics that allow for some formalized rules before you start

Virtue-based ethics

Positive ethics, morally admirable, observing good qualities and actions while remaining vigilant

Mean

another word for average

Median

The middlest number

Mode

The most popular number or the one that appears the most

Null Hypothese

No effect/ The explanation you are testing against, usually least interesting explanation (“There is no difference”)

Type I Error

You reject the null hypothesis but there is really no difference in the population parameter (false positive)

Type II Error

You fail to reject the null hypothesis, but there really is a difference between the population parameters (false negative)

Correlation

When you get 2 scores from one group

Regression

When predicting an outcome or score

Chi-square

Most appropriate use is with a single-factor, multiple group design in which the outcome measures are categorical and yield proportions (categorical data)

T-test

Mostly used with a single factor, two-group design to compare means (comparing 2 means)

ANOVA

Most appropriate to use with a single-factor, multiple-group design with an interval outcome measure (comparing 3 or more means)

Wilcoxon Signed-Ranks Test

Non-parametric test used for ordinal data or continuous data that doesn’t meet the assumptions of a parametric test (like a t-test)

• good for within-subjects

Latin Square

Matrix of n elements (letters) where each element appears exactly once in each column and exactly once in each row

Cronbach’s Alpha

Generalization of the Kuder-Richardson formula that estimates the average of all possible split-half reliability correlations when each test item has more than 2 responses

What is the replication crisis

Diederik Stapel published over 200 articles in prestigious social science journals, focusing on topics like stereotype threat and discrimination. When other laboratories were unable to replicate his findings, investigations revealed that Stapel had fabricated data in more than 30 peer-reviewed articles. This scandal prompted widespread reflection within psychology, as hundreds of classic studies weren't replicated and thousands of datasets were found to have been tampered with or analyzed dishonestly. Psychological research was set back by more than a decade and millions of dollars in research funding were wasted.

What is Dprime

If we plot the distribution of grades and take the average for each school, the difference of that average is our effect size, D prime. (it

Within-subjects

Same participants experience each condition/compare within an individual participant

• Each participant involved participates in ALL the conditions

Advantages/Disadvantages for Within-subjects

Advantages:

• Eliminates nearly all concerns of individual differences

• Requires fewer participants

• Greater statistical power

Disadvantages:

• Influence of different environmental variables (fatigue, weather, or gained knowledge)

• Participant attrition

Between-subjects

Different participants experience each condition/level of the IV

Compare across groups

Advantages/disadvantages of Between-subjects

Advantages:

• Measurement is uncontaminated by other treatment factors

• Eliminates risk of order effects

• Can be used for a wide variety of research questions

Disadvantages:

• Require more participants because each participant only contributes 1 score of data

• Individual differences: Scores obtained from different people, each with personal characteristics and are different from others

What are main effects?

The overall impact on one independent variable on the dependent variable, averaging across all levels of the other factor

• Describe general trends, but statistical tests determine if they are significant

What are interactions?

Occurs when the effect of one IV depends on the level of another IV, one factor changing or modifying how the other factor affects behaviour/performance

• Comparing the effect of different levels across factors

• May be visible of graphs as non-parallel lines

• When there is no ____ the effects of each factor are independent and consistent across all conditions

Level

How high or low it is on graph

Trend

Is it moving up or down between observations

Latency

When you start to see a big difference in results (almost always a bad thing)

Stability

• Deviation in responses

• Consistency of level and trend

Probability sampling

The size of the population is known before beginning so the odds of picking any person are known

Simple Random Sampling

Most simple approach to sampling

1. Clearly define population

2. List all members of population

3. Use a random process to get sample

systematic stratified sampling

For between group comparisons

1. Clearly define full population

2. List all members of population

3. Random process to pick starting point on list, then pick every fifth (*any number, this is just an example) participant down the list

Proportionate Stratified Sampling

Allows you to represent proportions

1. List all members of the population

2. Identify the sub-groups

3. Select the proper sized random samples from the sub-groups using simple random sampling

4. Combine sub-groups into sample

Cluster sampling

1. Clearly define population

2. Clearly define pre-existing clusters

3. List them the randomly sample them

Non-probability sampling

The size of the population is unknown/too big so the odds of picking any individual are unknown

Convenience sampling

• Most common method

Simply use the participants you have access to

No control so its representative of the population

Quota sampling

Recruit from a quota for the number of individuals per subgroup

Ex. Recruit first 15 from group 1, first 15 from group 2, first 15 from group 3, etc.