Exploratory Factor Analysis in Health and Human Services

¶ … 2008 Tenth Grade National-survey by Samhda/Icpsr

Substance Abuse & Mental Health Data

The study extracted data from the 2008 10th Grade National Survey SAMHDA/ICPSR. A five-factor exploratory factor analysis was conducted. Factors were labeled Marijuana, Alcohol, Weekend Alcohol, Graduation, and Periodicals. Total variance for the analysis was 88.467. The factor showing the highest percentage of variance was Marijuana. More variables loaded on Alcohol than other factors. Respondents did not perceive consumption of alcohol as high risk unless it was excessive, or excessive on weekends. Respondents reported the use of tobacco products as less indicative of risk compared to marijuana, which warranted higher disapproval. Overall, respondents consider themselves to be well informed and were confident about future opportunities for higher education.

An Exploratory-Factor Analysis of Selected Variables from the 2008 Tenth Grade National-Survey by SAMHDA/ICPSR

Access to national level behavioral health data is important to professionals in the fields of health and human service. Under the United States Department of Health and Human Services (HHS), the collection, analysis, and dissemination of behavioral health data is the primary responsibility of the Center for Behavioral Health Statistics and Quality (CBHSQ), Substance Abuse and Mental Health Services Administration (SAMHSA). The Substance Abuse and Mental Health Data Archive (SAMHDA) initiative is funded through a contract with SAMHSA. Moreover, the University of Michigan, Inter-University Consortium for Political and Social Research (ICPSR) is under contract to CBHSQ to disseminate data and maintain the SAMHDA website and the bibliography of publications. The purpose of this paper is to provide an exploratory factor analysis of selected variables from the 2008 10th Grade National Survey conducted SAMHDA and ICPSR.

Methods ?

The National Survey on Drug Use and Health (NSDUH) series is deigned to measure the prevalence and correlates of drug use in the United States. Surveys are conducted to obtain relevant data estimates on a quarterly and annual basis. Data is collected on use of controlled substances, such as drugs and alcohol, and illicit drugs by members of U.S. households who are 12 years of age or older. The sampling plan included stratification and weighting to assure representative samples across the states. Variance estimates were computed by using a clustered data analysis software package.

Participants

The respondent universe for the 2008 national survey was the civilian, noninstitutionalized population aged 12 years or older residing in the United States and the District of Columbia. The total targeted sample of 67,000 was allocated across three age groups: 12 to 17 years, 18 to 25 years, and 26 years and older. For the 2008 19th grade survey, the sampling frame would be those individuals in 10th grade from the 12 to 17-year-old group.

Data Collection?

The survey includes questions regarding respondents' age at first use, past month, annual, and lifetime use of eight drug classes. The survey also collects data about treatment history and perceived need for treatment for substance abuse. Questions regarding treatment for mental health disorders and questions from the Diagnostic and Statistical Manual (DSM) of Mental Disorders, which enable the application of diagnostic criteria, are included in the survey.

Data Analysis?

An exploratory factor analysis of select variables was conducted using Stata software.

Results

An exploratory factor analysis (EFA) was used to uncover the underlying structure of the large set of variables in the 2008 national survey. The objective was to identify relationships between measured variables for which no a priori hypotheses was formulated about the factors or the patterns of measured variables. Variables were selected that were represented by multiple measured variables in the analysis, with the common factors, unique factors, and errors of measurements expressing the measured variables. An underling assumption of exploratory factor analysis is that any measured variable may be associated with any factor. A rule of thumb was used to identify the number of variables as loading on factors if the absolute values of their factor loadings were l40 or greater. For this analysis, a five-factor structure was selected. The eigenvalues for the correlation matrix and the plot values from the largest to smallest were computed. The scree plot was used to determine the last substantial drop in the magnitude of eigenvalues, and to determine the number of factors to extract. The number of plotted points before the last drop was used as the number of factors to include in the model (see Figure 1).

Factor interpretation can be improved through rotation as it maximized the loading of each variable on one of the extracted factors while minimizing the loading on all other factors. The absolute values of the variables are changed through rotation, but the differential values are constant. The factors in this study were rotated orthogonally using the Varimax procedure with-Kaiser normalization.

Total variance for the analysis was 88.467. A single rotated factor loading with variable labels and eigenvalues, and the percentage of variance explained by the factors separately and cumulatively, is provided in Table 1

Factor 1.

Factor 1 shows eigenvalue of 3.585, % variance of 22.389, with five items loading on the factor. The factor 1 label is Marijuana.

Factor 2.

Factor 2 shows eigenvalue of 3.267, % variance of 20.408, with seven items loading on the factor. The factor 2 label is Alcohol.

Factor 3.

Factor 3 shows eigenvalue of 2.520, % variance of 15.743, with one item loading on the factor. The factor 3 label is Weekend Alcohol.

Factor 4.

Factor 4 shows eigenvalue of 2.484, % variance of 15.514, with four items loading on the factor. The factor 4 label is Graduation.

Factor 5.

Factor 5 shows eigenvalue of 3.585, % variance of 22.389, with four items loading on the factor. The factor 5 label is Periodicals.

Discussion

The use of marijuana and its attendant risks showed the strongest factor loading. Respondents' answers were associated with the language in the question items. For example, questions that used the words risk or disapprove showed high factor loading. The Alcohol and Weekend Alcohol variance showed the most unique pattern, with greater variance associated with alcohol use and less variance associated with disapproval of alcohol use.

The factor loading for Graduation suggests that the 10th graders are optimistic about graduation and attending college. The variance for Periodicals is high which indicates that the respondents consider their access to and use of periodicals such as newspapers and magazines to be high. Corresponding questions for TV and radio show low variance.