Chapter 7 Flashcards
(12 cards)
What is selection bias
Error that relates to the selection of participants
What is meausreable error
Error taht reflects the measurement of a study variable or variables
Will selection bias be reduce if you increase the sample size?
No, becuase the problem reflects systematic differecens.
What type of study is selection bias a concern in
case-control studies becuase the particpants are recruited in two seprate groups it is difficult to ensure the control group makes an appropriate comparrision to the cases
Specific causes of selection bias
Volunteers
Healthy Worker Effect
Low response Rates
Ascertainment of slection bias
Loss to follow up
V- ppl who vol differ from the average person
HWE - bias occurs due to ppl who work tend to be heaitler than ppl who don’t or can’t from disabilities
LRR - ppl who respond when response rates are low may be systematically different from pop
AoSB - comes up when biases are how participants are identified or included in study
Loss to follow up - they may be sicker or not motivated or busyier
To minizie selection biases researchers should 3 things
- have a clear definition of the population group they want to study (usally done)
- Have clearly defined eligility criteria to determine wheter specific individuals are included (usally done)
- recruit and samplpe participants so that they are representative of the population (rarely done)
What is random error
Random error refers to variability or fluctuations in data (measurements) that occur by chance rather than due to any specific factors or bias
What is systematic error
Is a consistent, reproducible error that skews measuremenets in a particular direction
What is a non-differential misclassification
When a measurement error and any resulting misclassifcation occur equally in all groups being comapred. They will ALWAYS lead to WEAKER observed associations than the truth
Differential misclassification
When the measurement error and misclassification occur to a greater extent in one group than another. Harder to predict than non-differentail misclassification
Measurement error - Recall bias
Occurs if cases are systematically more likely to overestimate or underestimate thier exposures than controls. If the cases feel guilty about it they would underreport exposure, if exposure if thought to cause disease cases overestimate exposure.
Measurement error - Interviewer or observer bias
Happens if data collectors ask questions in a different way for cases and controls.
