Group effects. or "effects".

Jul 22 2016 Published by under Replication, ReplicationCrisis

How many times do we see the publication of a group effect in an animal model that is really just a failure to replicate? Or a failure to completely replicate?

How many of those sex-differences, age-differences or strain-differences have been subjected to replication?

10 responses so far

  • bacillus says:

    I plan my expts so that regardless of gender all control mice die and all test mice live!

  • This is some wacked out incomprehensible jargon. Please define "group effect", "just a failure to replicate", "failure to completely replicate", "sex-differences", "strain-differences", "subjected to replication".

  • potnia theron says:

    I'm with CPP on this one. How about factors (fixed or random), treatments, levels, etc? Do you mean experimental factors of interest? "nuisance" factors? This is what partitioning of variation is for.

  • Zb says:

    I am highly suspect of subgroup effects and usually think they are post hoc analyses of experiments that did not show "interesting" results.

  • drugmonkey says:

    don't be intentionally stupid PP. I mean "hey, males and females are different on this outcome measure...PUBLISH". that is a group effect.

  • Grumble says:

    Huh? Are you saying that sometimes we compare two groups of animals we think are biologically different (male vs female, transgenic vs wt, etc) and when we get an effect in one group but not the other, the non-effect in the other group is really just a failure to replicate the effect in the one group? Because the different animals are really NOT different with respect to whatever it is that we are testing?

  • Dave says:

    Thats how i interpreted it Grumble.

  • physioprof says:

    DrugMonkey has a lot of trouble writing clear understandable English sentences. I think it's because he thinks everyone else can also hear the voices in his head.

  • Grumble says:

    Then the answers are many, many, and not enough.

    Why did you ask?

  • Another Assistant Prof says:

    I'd say the answer is statistically defined, if field-standard stat analyses were applied: 5%. That's the definition of the alpha value (for the most common tests, this would be the chance that you rejected your null by actual chance instead of detecting a true difference in the underlying populations means)

    But of course this makes all sorts of assumptions about normality, heteroscedasticity, etc., etc.

Leave a Reply