The Professional Counselor | Volume 12, Issue 2 141 Results Data Analysis Strategy Prior to conducting any analyses, the dataset was screened for data entry errors, unusual values, and extreme outliers; none were identified. Prior to running the negative binomial regression analysis, the categorical predictor variables (inferred binary gender, faculty rank) were dummy coded. All screening procedures and subsequent analyses were conducted using IBM SPSS (Version 28). To predict journal article publication counts, a negative binomial regression analysis was conducted because the criterion variable, journal article publications, represented a count variable that contained a large number of zero values and the variance of the distribution exhibited overdispersion (Fox, 2008). Power estimates for negative binomial regression models are less developed than those available for linear models. Nonetheless, traditional power estimates for general linear models (Cohen, 1988) and experimental estimates for generalized linear models (Doyle, 2009; Lyles et al., 2007) suggested that the negative binomial regression analysis likely had sufficient statistical power (> .80) to detect at least medium effect sizes. The following assumptions for negative binomial regression were examined: multicollinearity, residual plots, independence of residual errors, and the presence of any highly influential cases. No difficulties were identified. Ideally, a time series analysis is recommended for identifying trends or changes in longitudinal data across time (Yaffee & McGee, 2000). However, it is commonly recommended that a time series analysis should be based on a minimum of 50 observation periods (e.g., Tabachnick & Fidell, 2019). Power estimates for time series analyses can become very complex, and in some cases, 100 to 250 observational periods may be needed to reliably detect trends or seasonal patterns in time series data (Yaffee & McGee, 2000). It would not be feasible to track even a minimum of 50 years of journal article publications for a sizeable sample of counselor educators. Furthermore, inferential statistics— and accompanying power analyses—are needed for making inferences from a sample to the larger population from which the sample was drawn. Aside from inaccuracies on department websites, the counselor educators in this study represent the entire population of counselor educators at master’s-only programs in comprehensive universities who received their doctoral degrees at least 20 years ago. As Garson (2019) pointed out, “having data on all the cases in the population of interest eliminates the need for a random sample and, indeed, for significance testing at all” (p. 25). Consequently, the longitudinal analysis of this data will be limited to the creation and visual analysis of sequence charts. Characteristics of the Sample Regarding inferred binary gender, 51.9% (n = 81) of these counselor educators appeared to identify as female, and 48.1% (n = 75) appeared to identify as male. Two-thirds (n = 104, 66.7%) held the rank of full professor, and 33.3% (n = 52) held the rank of associate professor. The years in which they earned their terminal degrees ranged from 1970 to 2000 (Mdn = 1995.00, M = 1992.70, SD = 6.48). The number of years after earning their terminal degrees ranged from 20 to 50 (Mdn = 25.00, M = 27.30, SD = 6.48). Their terminal degrees included PhDs (n = 118, 75.6%), EdDs (n = 31, 19.9%), PsyDs (n = 4, 2.6%), and other (n = 3, 1.9%). Slightly over half of these faculty members had terminal degrees in counseling/ counselor education (n = 80, 51.3%), followed in frequency by counseling psychology, clinical psychology, or educational psychology (n = 47, 30.1%); education (n = 13, 8.3%); rehabilitation or rehabilitation psychology (n = 10, 6.4%); and other (n = 6, 3.8%). Almost two-thirds (n = 102, 65.4%) were faculty at public universities with the remainder (n = 54, 34.6%) being faculty at private universities. Regarding current Carnegie Classifications, over four-fifths were faculty at M1 institutions (n = 128, 82.1%), which was followed in frequency by M2 institutions (n = 20, 12.8%) and M3 institutions (n = 8, 5.1%).
RkJQdWJsaXNoZXIy NDU5MTM1