Chapter Nine Understanding Bias
The Nature of Bias Research bias may be thought of as a preference or predisposition to favour a particular outcome thus indicating a systematic distortion of research conclusions Typically the distortions are inadvertent, but they can also be intentional If bias is not addressed ina study, the reliability of findings is considered suspect
Bias in Quantitative Research Bias implies there is an unknown truth waiting to be described In quantitative designs the approach suggests that the researcher “knows” the truth & simply wants to confirm his/her knowledge eg. Questionnaire responses reflect the researcher’s previous knowledge, values, etc.
Bias in Qualitative Research In qualitative research the researcher is an integral part of the research design & the participant’s world To minimize bias in qualitative designs the researcher incorporates it into the design by: –bracketing –audit trail –selecting unfamiliar participants –selecting a topic that is not too close to the researcher on a personal level
Triple Biases: Nursing, Science, & Culture... We take our nursing predispositions with us to our research projects We tend to seek corroboration of our preconceptions, helps to make sense of a complicated world, reaffirms our pet theories Science itself potentially blinds researchers because of the expectations of findings, the belief in certain theories. Like culture, science produces blinders
Sexism: A Form of Bias Sexism is discrimination on the basis of gender RCT’s used mainly by medicine have made women victims of this approach, studying women mainly as objects, ignoring their needs & experiences
Types of Sexism in Research Androcentricity (male perspective)-based on individualism, materialism, & competitiveness - in contrast to the views of women, ethnic groups, & the poor who focus on family concerns rather than themselves Overgeneralization & overspecificity Gender Insensitivity Familism - treating family as unit of analysis, rather than the individual
Sources of Bias... need to design studies to systematically test alternative explanations researcher affect refers to the bias that results from a researcher having fallen in love with some pet theory or explanation
Bias: Selection of Problem Some things judged more important by funding agencies, one’s discipline peers bias is toward the conventional, standard projects & the selection of variables conventionally considered important & the exclusion of those conventionally considered unimportant probably still a bias toward quantitative approaches
Bias: Sampling Design results may be distorted by choosing to study sub-populations with known slants attitudes toward abortion in an urban community with a free standing abortion clinic vs rural communities bias is problematic in studies where the sample self-selects to participate
Bias: Funding SSHRC main funding for social science research special funding available in “hot” areas traditional areas better funding NSERC and CIHR are better funded research in a social context
Bias: Data Collection Experimenter effect: reference is to the influence of experimenter preferences and expectations Robert Rosenthal: the “smart rats” study. Clever Hans Expectancy Demand characteristics
Bias: Data Analysis Coding Errors –Random Error –Systematic Error Data Massaging Hunting
Bias: Reporting of Findings T.D. Sterling, 1959, Notes the problem of journals publishing only “statistically significant” findings.
Bias: Funding Possibility of funding agencies to determine what is important to know; have “they” got it right? Problems of emerging disciplines in competing with established ones
Advocacy Versus Pure Research P Mainstream research supportive of established interests in society. Is there a legitimate place to support the interests of minorities, women, people with disabilities, the working poor, the homeless…
Rules for Minimizing Bias Education Avoid Sexism Advocacy or Explanation? Descriptive Accuracy Let disconfirmation be your guide Policy Recommendations are Value Based Be skeptical of Research Findings
Rules Cont Read Literature Cautiously, Skeptically Distinguish Advocacy from Pure Research Use Theory to Generate Testable Hypotheses Be Sensitive to your own outcome preferences Do not disclose hypotheses to subjects or assistants
Rules Cont. Be Accepting of All Responses Specify Data Analysis Procedures in Advance Check for Random & Systematic Errors Report any Data Massaging