Establishing Relationships Flashcards
What are the different ways to talk about positive and negative relationships and what misunderstandings can arise from them?
- Two variable “go together”, as one goes up the other goes down, vice versa.
Problem: language can be confusing. It is not causal or a time relationship, more correlational. - Whenever H/L values on variables A are observed, H/L values on Variable B are observed.
More accurate way to talk about relationships without implying change over time - Positive or negative relationships between A and B.
Problems: this is a best guess, it is a predictions of missing data, not telling the future
How does one interpret a scatter plot and what tendency arises from estimating r from them?
Need two continuous variables
People underestimate the correlation coefficients
Correlation Coefficient of r:
-1: perfect negative relationship
0: no relationship
1: perfect positive relationship
Know how to interpret a contingency table and the relative frequency distributions
Values of variable x are the rows, values of variable y are the columns
Relative frequency distributions (rfd) must be different (these are the percentages) as is the best way to see if there is a relationship between two variables as raw numbers could be deceiving
Know about perfect and no relationships, effect size, and some common metrics including common language effect size
Perfect relationship and no relationship: knowing variable A perfect (or does not) tell us variable B. Incredibly rare and are commonly a fluke
Effect size: the absolute value of r (ignoring if it is positive or negative)
Common language effect size (CLES): best way to communicate the regular people about effect sizes
What is the best way to communicate effect sizes to people?
CLES (they are percentages!)
Be able to interpret numerical r and CLES in terms of small/medium/large
CLES based on r: average population with no effect size is 50%
Small: d = .2, r = .1, CLES = 53%
Medium: d = .5, r = .3, = CLES = 60%
Large: d = .8, r = .5, CLES = 67%
How are effect sizes often misinterpreted when psychology research is communicated to the public? What is the advantage of expressing them in natural units?
People often misinterpret effect size, think it is larger than it is
Best to find out an effect size in natural units. Such as percentages and regular numbers
Know how to define and identify single-cell, single-diagonal, and single-row errors, making the case for a full set of data to make decisions. What are some of the likely causes of these errors, psychologically?
Single cell error: error based on one instance of something (such as testimonials, egocentric bias, case studies)
Made up by advertisers, paid endorsements, reasoning from ones own perspective
Single-Diagonal: like single cell, but you add another cell that confirms your hypothesis
Made up by assuming all relationships are perfect
Single row error: errors from having data in one row or column of the contingency table but not the other.
Superstitious behaviour is from this
e.g. top row filled out but bottom is not
What is base rate error? How does it show th need to look at both raw numbers and percentages?
Base rate error: when you have all the percentages by row but not the raw numbers
What is Goodhart’s law?
If sufficient rewards are attached to some measure, people will game the system to increase the scores, leading to the measure not being an effective measurement anymore.
How are statistics misapplied in practice according to Goodhart’s law?
Science citation metrics, editors pressure authors to reference studies from the same journal. Some publish an excess of papers in January to get the most citations.
What is mathiness?
Refers to formulas and expressions that may look and feel like math, even as they disregard actual mathematics.
How is mathiness used to promote bogus formulas and equations in the press?
News outlets report on bogus formulas, using these to promote flawed narratives, conclusions, by attempting to come across as more valid through the use of math
What is a zombie statsitic?
Numbers that are citied badly out of context, are outdated or were made up. But they are quoted so often they will not die
What do zombie statistics say about the relationships between news media and research?
Even though research can attempt to counter the claims made in media, it does not stop the spread of it, even when disproven.
