Research design Henry Flashcards
what si the power of a statistical test?
The power of a test is the probability to reject H0 if H1 is true.
In albatross example: Power = probability to correctly conclude from your sample that the mean duration of the trip differs between the 2 treatments (magnet, fake magnet)
To be able to carry out the power analysis in G*Power the following information is required:
- The statistical test you
are you going to use. - The desired significance
level α (mostly 5%)
=probability to reject H0 if
H0 is true - The desired Power 1-β
(often in the range 60-
80%) = probability to
reject H0 if H1 is true - The sample standard
deviation of the variable
(from a pilot study or
from literature) - The desired change to be
detected by the statistical
test with the desired
Power - If you test one- of two
sided - Allocation ratio (n2/n1
=sample size group
magnet /sample size
group fake magnet)
A true experiment is when:
- There is a control group
- Animals are randomly
assigned to the
treatment or control
group (each animal has t
he same chance of
receiving any one
treatment)
research design:
de wijze waarop de experimentele eenheden (vaak dieren) gerangschikt zijn m.b.t. de behandelingen en de mogelijke storende invloeden.
Experimental design can be split into two categories:
- Between subject designs
- Within-subject designs
Within-subject designs:
- Each experimental unit
undergoes all the
treatments - Between-subject variation
is negligible because each
animal functions as its
own control
between-subject designs
- Each experimental unit
undergoes one treatment - All animals receive the
treatment at the same
time – time-effects are
negligible
voordelen between-subject designs
- It’s a simple research
design - The statistical tests to be
used are simple and can
also be used when the
number of animals per
group is not equal
(Albatross example) - Every animal is only
treated once, resulting in
a “short” experiment:
the animal suffers less
smaller chance that an
animal must be removed
from the experiment
prematurely
nadeel between-subject designs
Each treatment is applied to a different set of animals. So, differences between the two treatments can also be due to other factors than the treatment itself because of between-subject variations (e.g. age or sex).
block factor
A block-factor is a factor in which we have no interest in but which we take into consideration in the research design because it influences the dependent variable. Its a factor that is measured at the nominal or ordinal level
covariate
A covariate is a factor in which we have no interest in but which we take into consideration in the research design because it influences the dependent variable. Its a factor that is measured at the ratio or interval level
voordelen within-subject designs:
- Er wordt niet
gecorrigeerd voor
tijdseffecten, er loopt
namelijk geen
controlegroep mee
gedurende de hele proef.
(zie 10.3.1) - Behandelingseffect moet
reversibel zijn (dat wil
zeggen dat het dier
terugkeert naar de
toestand die hij had voor
de behandeling ) en het
dier mag niet gedood
worden - Er wordt dus niet
gecorrigeerd voor
zogenaamde carry-over
effecten (zie 10.3.2)
voordelen within-subject designs
Controle voor diereffecten; minder dieren nodig (hogere power)
wat betekent counter-balancing?
Er wordt gecorrigeerd voor tijdseffecten omdat elke behandeling in elke periode wordt gegeven.
Door de volgorde van taken te balanceren (bijvoorbeeld sommige deelnemers beginnen met taak A en andere met taak B).
carry-over effect
occur when a treatment contineus to affect the subsequent state of experimental subjects.
washout period
A period of normal conditions between the two treatments, to minimize carry-over effects.
Factoriele schema’s:
worden gebruikt als je meerdere behandelingen tegelijk wil testen. Een dier krijgt dus twee of meer behandelingen tegelijkertijd.
