Psychologists around the world may be making conclusions from faulty assessments according to a professor visiting the UCSB Methods for Social Sciences (QMSS) Colloquium.
Trevor Bond from James Cook University in Australia gave a talk on Thursday in North Hall about what he said were the shortcomings of psychological testing and measurement. One of the biggest problems with getting results from current psychological testing is the qualitative nature of testing methods. Bond proposed the use of a method that would add objectivity to the field and perhaps standardize psychological testing.
In his lecture entitled “Have psychologists been measuring anything at all? Rasch measurement and the science of psychology,” Bond proposed that the Rasch Model would allow more quantitative and stable conclusions to be made from tests. These models are already starting to be used in psychology, education and rehabilitation medicine.
The Rasch Model is named after Georg Rasch who originally published his method in 1960. Even though over 40 years have passed, the method is not used very often in current research. Bond said this is because there are millions of dollars invested in other methods that have become conventional. He said researchers are often extremely attached to old methods even though they may not be the best.
“At the start of the new millennium, there is enough evidence to argue that psychoanalysis has not been measuring anything at all,” Bond said. “We’ve been doing the psycho but not the metrics in psychometrics.”
Bond said he is dissatisfied with current testing methods because a score may be composed of added numbers that cannot be properly added. An example would be a test in which the score on each question is added to produce a score for the whole test.
“It is not good enough to assign numbers to behavior and call this measurement,” Bond said. “Psychology should be a quantitative, rational science.”
On an anger assessment test, a response of “strongly agree” may be worth five points, while a score of “strongly disagree” would be worth zero points. One problem with this approach is that each question contributes to the overall score in an equal way, although one question may concern hitting other people while another may ask about breaking pencils, Bond said.
In contrast, the Rasch Model would assign a weight to each question, to be determined by the percentage of people that answered “strongly agree” on previous tests. After compiling many tests it could be determined which questions were least likely to receive the “strongly agree” response. These are the questions that would be weighted the most influential on future tests because so few people tended to respond “strongly agree.” This process is called conjoint measurement.
“[Other researchers] have outlined the principles and properties of conjoint measurement that would bring the same sort of rigorous measures to human sciences that the physical sciences have enjoyed for a considerable time,” Bond said.
Professor of Education Michael Furlong listened to Bond’s talk and said he felt that the Rasch Model would greatly improve accuracy in testing.
“I am interested in child assessment, and there are a lot of variables that are important to understand,” Furlong said. “This is the most enthralling lecture I’ve heard in psychology.”
Specific questions are often written to assess different aspects of anger such as verbal or physical.