Upgrade to remove ads
Lecture 7 - Research integrity
Terms in this set (30)
What are the four hoursemen of the astatalypse?
Low power, p-hacking, harking, and publication bias
What is p-hacking
Messing with data to improve p-value
Hypothesising after results are known
Not publishing null results
Titles can be...
Misleading, imply significant results when not
________ can be missed from graphs to give the illusion of significant effects
Methodological issues include...
Gender imbalance, low sample size, inappropriate age range
What are median splits and what is the problem with them?
Creating arbitrary dichotomous categories out of the median, implies everyone in each category is the same and everyone in different categories are vastly different
Post-hoc tests are done...
Without justification, NS
The worst form of statistical malpractice is...
Fabrication of data
What implies that malpractice is common?
Most research is under-powered yet significant
______ prestegious journals are _____ likely to have replicated results, and _____ retractions
More, less, more
Studies are likely to be "file drawered" when what happens?
Many published studies have significance of...
P-hacking is common in which analysis?
Information omission (e.g. existence of other variables), controlling for many variables, changing the DV, and analysing part way through then stopping at significance are a part of which malpractice?
What should you do about outliers?
Establish treatment of them before data collection
Using multiple analyses and using different methods to test the same thing are examples of what?
Which influential trial was an example of changing the DV?
The PACE trial
What is null hacking?
Malpractice and bad design to make results non-significant
Who is likely to do null hacking
How can group design be a problem?
No control groups or non-equivalent control groups
What is preregistration?
Submitting research plans before carrying out the study
What are Bayes classifications?
More stringent p-value criteria to counteract low power
What level does Bayes classification suggest?
Why are Bayes classifications a bad solution?
P-hacking is still possible, more costly to get significant results, increases false negatives, harder to replicate
What should always be reported to help statistical clarity?
Effect sizes, confidence intervals
What should be done instead of an independent samples t-test or univariate ANOVA?
How are unplanned analyses reported in pre-registration?
Reported as exploratory analyses.
What questions are answered in pre-registration?
Hypothesis, IVs, no of conditions pps assigned to, analysis, secondary analyses, sample size, 'other', and whether any data already been collected.
THIS SET IS OFTEN IN FOLDERS WITH...
Lectures 4 and 5 - reliability and validity (scale…
Lecture 6 - Principle Component Analysis (PCA)
Lecture 8 - MANOVA
Lecture 9 - ANCOVA
YOU MIGHT ALSO LIKE...
Research methods - quantitative
MKT 473 Exam 4- Franke
OTHER SETS BY THIS CREATOR
Lecture 1 - introduction
Lecture 7 - Word learning
Lecture 9 - Hunger and satiety III: Central factor…
Lecture 7 part 2 - hunger and satiety 1: periphera…