This paper explores the extent of redundancy among the small area indicators of the 1970 Mental Health Demographic Profile System (MHDPS). Redundancy is equated with interindicator Pearson correlations equal to or greater than .80 for the census tracts within six metropolitan counties. The analysis was undertaken because the assertion had been made that the association among indicators was high, particularly within major social area dimensions, and consequently, a few indicators could be used to provide the information contained in the system. The analysis failed to support this thesis. The paper concludes by making a case for a judgmental selection of items and briefly reviewing the 1980 MHDPS.
The Mental Health Demographic Profile System (MHDPS), a general purpose data base system containing 1970 census data, was established by the National Institute of Mental Health (NIMH) to assist planning and administration by providing small area social indicators for need assessment and program evaluation. Since its establishment, the 1970 system has been used extensively by mental health and health administrators for program planning. While no systematic documentation of compliments and complaints was made, there was a growing sense among the NIMH staff, as they reviewed this system of census indicators in preparation for its 1980 update, that changes in the system were required. While generally highly valued among its varied users, there was a concern that MHDPS presented its original administrative users with more information than they could comfortably and effectively utilize. Although some consensus appeared to emerge from user comments indicating a desire for a few nonredundant indicators, no consensus emerged indicating which of the approximately 160 they should be. Some researchers interested in using linear techniques, noting potential multicollinearity among the indicators, had asserted that only a limited number of indicators were really necessary to account for variation in the data set and that a significant number of items were so highly correlated within mental health service areas or counties as to be unnecessary regardless of their substantive meaning. These remarks suggested a way of reviewing the data set and possibly reducing it for both administrative and scientific users. Unnecessary items containing overlapping information might be indicated by high interitem correlations. Thus, these comments became linked in the minds of the staff under the general term redundancy.
This paper reports the initial examination of the 1970 MHDPS for redundancy or high interindicator correlations.
1970 Mental Health Demographic Profile System
MHDPS contains items abstracted from the 1970 Census of Population and Housing. These items represent social area dimensions (socioeconomic status, family status,