The article “Self-neglect and cognitive function among community-dwelling older persons” by Dong, Wilson, Mendes de Leon, and Evans (2010) reports an epidemiological study of associations of self-neglect among the elderly population to a number of factors. To test these associations, a number of linear regressions were conducted. While the choice of the test was appropriate (Forthofer, Lee, & Hernandez, 2007), the article has a number of issues related to reporting the problems. However, the conclusions of the article seem to be appropriate.
Appropriateness of the Selected Test
It is apparent that linear regressions were a proper choice of analysis in this case. Multiple regressions are utilized to predict the values of a single quantitative dependent variable from a number of independent variables (which can be either quantitative or categorical) (Warner, 2013). In this case, the authors attempted to predict the levels of self-neglect (measured on a quantitative scale) from the levels of global cognitive performance, perceptual speed, and episodic memory (Dong et al., 2010), so using linear regressions was correct.
Problems with Reported Results
Apparently, there are some problems with the way in which the results are provided. First, the subsection “Metric properties of self-neglect measure” reports some relationship between a number of variables. For instance, it is stated that “higher Self-Neglect Scale score was associated with older age (r=.012, p<0.001) and female gender (r=-0.07, p=0.03)” (Dong et al., 2010, p. 801), but it is unclear which statistics are reported. One might assume that if associations are mentioned, then r stands for Pearson’s correlation coefficient. However, in this case, these Pearson’s rs are too close to 0 to speak about a correlation, which makes it difficult to interpret this subsection.
Another issue is the way in which the results of regressions are reported. For example, it is stated that linear regression models were built to “estimate the association between self-neglect and cognitive function” (Dong et al., 2010, p. 801), and somewhat later their results are reported: “being reported for self-neglect was associated with lower levels of global cognitive function” (Dong et al., 2010, p. 801). This wording appears doubtful, for linear regressions allow for predicting change in the dependent variable for one unit of change in an independent variable (Laerd Statistics, 2013), and this wording makes it seem as if self-neglect was the independent variable in this case, which appears wrong. Furthermore, the authors speak about an association, which seems doubtful in the context of a linear regression, where it is possible to speak about predicting variables.
A problem exists on p. 802, where the authors confused numbers of tables: “self-neglect severity had a marginally significant association with MMSE (Table 2, model A)” (Dong et al., 2010, p. 802), but table 2 reports different things.
It is stated that for a multiple regression, at least the βs, their confidence intervals, p-values, as well as some general description of the model (e.g., the values of R2) should be reported (Field, 2012). Dong et al. (2010) report parameter estimates, their standard errors (which is enough for calculating their confidence intervals), p-values, and R2 values (p. 801-802); apparently, this meets the minimum outlined by Field (2013) if the “parameter estimates” are βs.
Despite the described problems, it is possible to mostly accept the conclusions made by the authors if one assumes that the analysis was conducted properly. Tests other than linear regressions should not be used for this data, because, as was explained before, the choice of the test is appropriate. However, it would help if the authors added more clarity to their report. The interpretation of the results by the authors, on the other hand, seems appropriate.
Dong, X., Wilson, R. S., Mendes de Leon, C. F., & Evans, D. A. (2010). Self-neglect and cognitive function among community-dwelling older persons. International Journal of Geriatric Psychiatry, 25(8), 798-806. Web.
Field, A. (2013). Discovering statistics using IBM SPSS Statistics (4th ed.). Thousand Oaks, CA: SAGE Publications.
Forthofer, R. N., Lee, E. S., & Hernandez, M. (2007). Biostatistics: A guide to design, analysis, and discovery (2nd ed.). Burlington, MA: Elsevier Academic Press.
Laerd Statistics. (2013). Multiple regression analysis using SPSS Statistics. Web.
Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA: SAGE Publications.