CHAPTER 10. RESPONSE BIASES 1
Response Bias
Response cannot be interpreted validly, in terms of the
psychological attribute (we are measuring something else)
Bias is often systematic (part of the true score variance) Sxt2
and not in Sxe2
We can see an increase reliability as a consequence of
(systematic) bias
Influenced by response sets and response styles
Types of Response Biases
Acquiescence Bias: you give the same response all time
regardless of the meaning of the statements. (yes-saying
bias, nay-saying bias)
Difficult to determine which respondents have high level of
construct from the ones that respond with acquiescent bias.
Occurs when complex items, testing situation with
distractions, difficulty of respondents to understand the
material.
o Leads to overestimation of reliability when internal
consistency measures are used (because of systematic
variance)
o Correlation of the construct is also (over)estimated.
o Removing the biased participant > correlation corrected
(smaller)
Extreme and Moderate Responding: tendency to use or
avoid extremes- extreme/ moderate answers; hard to
determine if it is due to construct or due to bias.
o People with identical levels on construct > differ due to
their answer tendency
o People with difference in the construct levels > do not
vary because of their tendency
o We can remove the middle category (not ideal), add more
categories e.g. in some cases group categories, ppl that
score on extremes more due to bias
, CHAPTER 10. RESPONSE BIASES 2
o Leads to distortions of the correlation (higher than in
reality)
Response Bias
Response cannot be interpreted validly, in terms of the
psychological attribute (we are measuring something else)
Bias is often systematic (part of the true score variance) Sxt2
and not in Sxe2
We can see an increase reliability as a consequence of
(systematic) bias
Influenced by response sets and response styles
Types of Response Biases
Acquiescence Bias: you give the same response all time
regardless of the meaning of the statements. (yes-saying
bias, nay-saying bias)
Difficult to determine which respondents have high level of
construct from the ones that respond with acquiescent bias.
Occurs when complex items, testing situation with
distractions, difficulty of respondents to understand the
material.
o Leads to overestimation of reliability when internal
consistency measures are used (because of systematic
variance)
o Correlation of the construct is also (over)estimated.
o Removing the biased participant > correlation corrected
(smaller)
Extreme and Moderate Responding: tendency to use or
avoid extremes- extreme/ moderate answers; hard to
determine if it is due to construct or due to bias.
o People with identical levels on construct > differ due to
their answer tendency
o People with difference in the construct levels > do not
vary because of their tendency
o We can remove the middle category (not ideal), add more
categories e.g. in some cases group categories, ppl that
score on extremes more due to bias
, CHAPTER 10. RESPONSE BIASES 2
o Leads to distortions of the correlation (higher than in
reality)
