exam 2 experimental design

define confounding

when the effect of 2 or more variables on the response variable cannot be distinguished from eachother

what is the problem with confounded results

• can cause false relationships

• can mask real relationships

• you don’t know what caused the results

how do you know if study results are confounded

• think about what the response variables are actually measuring

• ask yourself if there is a variable other than the original that will impact the response variable

• if so, is the variable controlled for

how to avoid confounding results

identify external variables that could impact the response variable and then block for those

what is assessment bias

bias in evaluating the response to treatments

when is assessment bias likely

when treatments are known & assessment is subjective

• subjected = ascaled, categorical

examples of response variables that would likely have assessment bias if the treatment were known

• body condition score

• lameness score

• subjective catagorical variables

how to reduce assessment bias

• double blind studies

• single blind studies

double blind study

removes any tendency for assessors to favor or disfavor a particular treatment (particularily treatment for subjective measurements)

single blind study

either the assessor of the response to the treatments or the care-takers of the animals do not know which treatment each animal is receiving

benefit of a double blind study

it removes any tendency for you to favor/disfavor a particular treatment and removes any tendency for caretakers to handle animals differently

why is it important to treat all subjects alike

it can affect the study results

how does double blind studies differ for human and animal studies

• neither the assessor of the response to the treatments nor the subjects receiving a treatment know which treatment a subject is receiving

• doctors or other medical staff may treat patients or make assessment differently if they have knowledge of the treatment a subject is receiving

• if a patient knows they are receiving a placebo, it may decrease the placebo effect & bias experiments in favor of other treatment

how does single blind studies differ for human and animal studies

either the assessor of the response to the treatments or the person getting the treatment does not know which treatment a subject is receiving

how does variability affect the ability to detect treatment differences if they exist

makes it more difficult

can we eliminate all variation in an experiment animal study

no, it wouldn’t be practical

how does a blocking factor help you to reduce the impact of animal variability on your ability to detect treatment differences (if they exist)?

• it allows you to separate the effect of block from the effect of treatment

• is a method to control variation between animals in your study

what are methods to reduce animal variation in an experimental study

• self pairing

• natural pairing

self pairing

• when an animal serves as their own control

• study designs: latin square & switchback design

natural pairing

data from a study where each observation in one sample is naturally linked to an observation in a second sample because they come from the same individual or related units

what is natural pairing used for

used to reduce variability by controlling for extraneous factors, such as comparing a subject’s measurements before and after a treatment or pairing twins for genetic studies

• ex: littermates