Statistics for the Behavioral Sciences Generating Research Questions

Statistics for the Behavioral Sciences Generating Research Questions HDFS 7050 - Bubb, R. 1

Research Questions • Start Broadly • What are you generally interested in • Write down all of your ideas • Consider various hypotheses • Get Specific • Define your outcomes and predictors • Think about how you would measure them • Ask Experts and ask experts • Conduct a literature review (unpublished also) HDFS 7050 - Bubb, R. 2

Research Questions • What is your target population? • Children, adolescents, adults, aging adults, parents, teachers, policy makers • Income levels, rural, urban, education levels • What is the level of aggregation? • Individual, family, community, state, county • Maybe multiple levels HDFS 7050 - Bubb, R. 3

Research Questions • What is the outcome variable going to be? • This is the variable of interest • Achievement, child nutrition, social interactions • What are the predictor variables going to be? • This is the variable that is related to systematic variation in the outcome (explains outcome responses) • Age, access to fresh produce, number of friends • Do predictor variables interact? HDFS 7050 - Bubb, R. 4

Research Questions • Some predictor variables can be influenced by the researcher • Whether participants receive an intervention • This is called researcher manipulation • If a manipulation is randomly assigned then • It can be assumed the groups are equivalent • Therefore all other possible explanations for the findings (due to individual differences) should be eliminated • This is an experimental design HDFS 7050 - Bubb, R. 5

Operational Definitions • Operational definitions should: • Include the units used in the measurement • Identify the procedures needed to the collect data • Be specific to allow for easy replication

Operational Definitions • Raises challenges for measurement and data collection: • Do instruments already exist or do you need to make them? • How will they be administered? Face-to-face, by e-mail, by phone? • Are multiple measures of the same construct needed? • Are you restricting the range of variability?

Research Questions • What are some covariate variables? • Related to outcome but not the predictor • Literature review is important here • Gender, ethnicity, SES, neighborhood • What are potential confounding variables? • Related to both the predictor and outcome • Cannot be disentangled through statistical analyses • Account for in the research design HDFS 7050 - Bubb, R. 8

Research Questions Is “question predictor” related to “outcome variable” among “target population, ” accounting for “covariate variables”? What is it we are trying to understand? What is it we think might explain differences in our outcome? Who do you want to study? HDFS 7050 - Bubb, R. Rule out other explanatory variables 9

Research Questions Is poor physical health associated with lower reading and mathematics achievement among economically vulnerable children in rural Alabama, accounting for child gender and family SES? ” Outcome variables Key question predictor Target population Covariate variables or rival hypotheses HDFS 7050 - Bubb, R. 10