Appendix B: Detailed Methology AARP Multicultural Survey
B. Statistical Methods
In survey research, statistics serve two essential functions. The first role is to provide a method for organizing, summarizing, and communicating the data that represent the opinions, attitudes, behaviors, and personal characteristics measured in the surveys. The set of mathematical procedures serving this function is called descriptive statistics because it characterizes the basic nature of the data. Frequencies, percentages, and means or averages are mathematical methods for effectively organizing, displaying, and communicating survey data recognizable to most people.
The second function, inferential statistics, is a body of mathematical procedures for arriving at conclusions advancing beyond the description of the survey data to the relationship among them. One group of procedures displays and measures the relation among two variables and the significance of the measures of associations. For example, common to most users of survey research is bivariate statistics such as cross tabulation—the display of data from two categorical variables in a simple tabular format with tests of significance (e.g., chi squares) and a simple regression—the measure of the linear relationship between two quantitative variables that can be depicted in a graph with tests of significance (e.g., correlations).
Somewhat less common to survey research users is multivariate statistics such as multiple linear regression and factor analysis. There are often several variables and sometimes many variables each of which may contribute to the understanding of the relation among the variables in the survey data. These statistical procedures help to tease out such relationships and associations. Because these techniques were used to guide the analysis of the data in this survey project, a brief explanation of the ones used and a description of how they were applied to the data and used in the analysis are presented below.
Tables were created displaying the relationship between a set of independent variables, including demographic items (gender, age, education, income, race/ethnicity, place of birth, employment, marital status, region, and urban/suburban/rural area), personal circumstances (health, satisfaction with life, family outlook, children, availability of flex time for family care, and who raised the respondent), and family circumstances (being part of the sandwiched generation – having children under 21 and parents or in-laws living, who respondent considers to be a family member, the composition of the respondent’s household, and one’s care-giver status), and all the variables (all individual questions were treated as dependent variables) in the study.
These tables were run against the whole weighted sample providing nationally representative data on the relationships between the independent and dependent variables. Due to the considerable interest in how these relationships might vary by race or ethnicity, the complete set of tables was also run for the four major race/ethnic groups in the study: non-Hispanic whites; African-Americans; Hispanics; and Asian Americans. This provided an instant control for