Mystery Category!
To prevent a null result from being "uninterpretable," researchers must use these types of checks to prove a treatment actually reached the participants.
manipulation checks
When a confound is present, this specific type of validity—the ability to establish a causal relationship—is lost.
internal validity
This common mantra reminds researchers that a p-value > .05 does not prove an effect is zero.
absence of evidence is not evidence of absence
while this process "simply works," its failure is usually a byproduct of an insufficient sample size rather than a procedural error.
random assignment
A confound may lead to this specific statistical error, where a difference between groups is observed but is actually due to the confound rather than the manipulation.
spurious correlation
If a new intervention shows no effect but this "known quantity" included in the study does, the null result is considered highly credible.
What is a positive control
This "procedural fix" for confounds requires being intentional and thinking about the experiment from as many angles as possible.
good experimental design
Beyond just "experimental tidiness," confounds are dangerous because they can lead to these results, causing researchers to "chase illusions."
false positives
This controversial practice involves running statistical tests on demographics to ensure "equality" between conditions, though the logic is often flawed.
checking randomization
when confounds "fundamentally mislead" a study, the observed effects will likely fail to do this.
generalize