Mary G. Lieberman’s research while affiliated with Florida Atlantic University and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (41)


Leverage the Literature Review for Statistical Comparisons
  • Conference Paper
  • Full-text available

April 2024

·

212 Reads

·

Mary G Lieberman

·

This is a suggestion that we encourage examination of the fit of a study sample to literature review samples in respect to mean and variance to facilitate integration. In addition, a suggestion for how this might be done is offered. An added burden is that this must be done from aggregate data. Some detailing of popular software alternatives is given and software that will easily accomplish all of this with simple entry of these aggregate data is demonstrated and provided. An evaluation of the “importance to the researcher” of this technique by a group of doctoral students was quite positive (p < .001, ES=1.10).

Download

Assessing Content Validity: An Extension to Lawshe's Method

This paper extends Lawshe’s (1975) method of calculating indices and inferential tests for candidate items’ content validity from judges’ ratings. The extension includes incorporation of the calculations into software that have heretofore necessitated tables for the original dichotomous scoring and extends that binomial method for dichotomous judges’ decisions to a multinomial method for ordinal decisions. Item disagreement is also covered. An example is provided with an illustration of the increased statistical power afforded by using the multinomial referent. An easy-to-use Excel spreadsheet that accomplishes all calculations from the binomial or multinomial perspective is offered.


Handling Covariate Patterns in the Examination of Logistic Regression Residuals

Logistic Regression influence and leverage residual diagnostics must be calculated by covariate pattern rather than by subject. As this is too often violated, paths to this correct analysis are discussed in respect to different software packages. Therein, a software requirement for correct analysis of some packages (SAS, Minitab, R) is to first convert raw data into a compressed Event/Trial format. Stata compresses data into covariate patterns internally doing the analysis correctly without altered input. As SPSS calculates diagnostics by case, not offering Event/Trial format for logistic regression, there is no way within SPSS to get appropriate diagnostics for covariate patterns with more than one case. As compressing data to Event/Trial format manually is tedious and error prone, an Excel spreadsheet is offered that creates a copy and paste Event/Trial formatted data file automatically for that purpose. If SPSS is used, a simple copy-and-paste of predicted probabilities into the program renders the diagnostics supplied by SPSS correctly by covariate pattern, as well as many diagnostics not supplied by SPSS. In addition, as SPSS does not supply typical logistic regression diagnostic plots supported in other programs, those are also rendered. An example is provided to illustrate these options.


Simultaneously Maximizing Coefficient Alpha and Minimizing its Standard Error

May 2023

·

42 Reads

·

2 Citations

The purpose of this paper is to provide a procedure for maximizing coefficient α and minimizing its standard error among all possible subsets of items in a survey, or more generally, any objectively scored instrument. The option of forcing items into subsets to preserve validity is also supported. An Excel spreadsheet was created to execute the large amount of combinatoric calculations and is freely available. The results are confidence intervals for α across all subsets of items that give expectation limits for sampling both items and subjects.


Responsible Partialling

Partialling correlations (partial, semipartial or bipartial) is built upon regression equation(s). It is imperative that the assumptions of those models be considered. Those assumptions rest on the OLS residuals of those models (e.g., linearity, homoscedasticity, normality, independence, absence from data with extreme influence). Partial correlations of all types depend on residuals, but consideration of fit of data in respect to those residuals is seldom done. Although this can be accomplished with repeated subsequent analyses, partialling in commercial software does not directly render these assumption diagnostics. As these are mandatory for use of least squares, examples are produced and an Excel program that automatically performs an exhaustive set of diagnostic analyses and plots for partialling is offered.


Experimenting With Least Squares, Curvilinearity, Leverage, and Influence

To assist our students in grasping the notion of least-squares, and the effects of curvilinearity, leverage, and influence, we developed a hands-on lab using custom software for the bivariate case. This is administered early in a course covering regression as well as other topics and is based on the concept that experimentation with raw data will facilitate awareness and enhance comprehension and retention of these important aspects of regression modeling. This software poses as a game in which students successively guess slopes based on the standardized scatterplot and a plot of MSE. Through sufficient “dwell-time,” students learn first-hand that r minimizes MSE. In addition, the software automatically flags points with leverage (using the hat diagonal) and influence (using Cook’s d). The inability of least squares to handle curvilinear models is also easily demonstrated. Evaluation of learning by students was very positive. Examples are provided and a secure link for software download is included.






Citations (17)


... Serlin and Kaiser (1976) suggested removing items that have low weights derived from the first principal component of the intercorrelation matrix of items. Examination of the alpha of all possible subsets of items has also been suggested (Morris, 1978a(Morris, , 1978bLieberman, Morris, & Vásquez-Colina, 2023). The latter can be computationally intensive; with k items, alphas for 2 k -1 -k combination of items are necessary. ...

Reference:

Likert Leftovers: A Case for Attending to Items Deleted to Achieve Internal Consistency
Simultaneously Maximizing Coefficient Alpha and Minimizing its Standard Error

... Morris, Florida Atlantic University, Lieberman, and Florida Atlantic University [22] employed a multiple regression model to determine that attitudes were the primary determinant of an intention to wear seat belts in scenarios of safe driving, while norms were the primary determinant of behavior in hazardous driving situations. In order to examine this phenomenon, Fishman, Yang, and Mandell [23] carried out a series of three studies with the aim of altering the attitudes of participants regarding the significance of utilizing a seat belt while operating a vehicle. ...

Integration of ANOVA and Multiple Regression for Beginning Statistics Students
  • Citing Article
  • January 2023

General Linear Model Journal

... This then allows delivery of a statistical methods course, from beginning to advanced using only software that is local and familiar to the student. We offered this software in the height of the pandemic (Morris & Lieberman, 2021). We offer many extensions, our experiences, and student evaluations here. ...

Pandemic Package for the Social Sciences -- An Ongoing Effort Toward Access and Accuracy

... In order to assess symmetry and disagreement, Morris and Lieberman's (2018) method was applied. This method conducts the Cooper (1976) and Whitney (1978) inferential tests for polarity of sentiment which depend on the normal and t distributions respectively for larger samples, providing effect size estimates as well. ...

Treatment of single Likert-type items in surveys

... Furthermore, we performed a detailed correlation analysis, employing cross-correlation to gauge the importance of features for classification: a relevant feature has a high correlation with the target variable but a low correlation with other features [30]. The analysis also shows the potential impact of multicollinearity on the classification of this dataset [31] by calculating the level of the features' mutual correlation. Lastly, cross-correlation has been Sensors 2025, 25, 1468 6 of 20 used to detect data overlap by showing the level of similarity between the time series from different classes. ...

The Precise Effect of Multicollinearity on Classification Prediction

... This study uses correlation coefficient > 0.80 to indicate collinearity and multicollinearity between independent variables as the sample number is small (Berry & Feldman, 1985). VIF more than 2 is also used to specify multicollinearity after multivariate analysis (Morris & Lieberman, 2012;Tabachnick & Fidell, 2007), particularly in PLS-SEM (see Chapter 7). ...

Selecting a Two-Group Classification Weighting Algorithm: Take Two

... Some studies have shown that adult adoptees are at higher risk of depression (Borders et al. 2000;Cubito and Brandon 2000), low selfesteem (Borders et al. 2000;Levy-Shiff 2001), substance abuse (Bohman and von Knorring 1979), personality disorders (Bohman and von Knorring 1979), alienation (Lieberman and Morris 2004), psychological distress ( Levy-Shiff 2001;Smyer et al. 1998), low educational achievement, and low IQ ( Teasdale and Owen 1986). Other studies have found no difference in substance abuse or criminality (Bohman and Sigvardsson 1990;Borders et al. 2000), life satisfaction ( Borders et al. 2000), or general psychological adjustment ( Collishaw et al. 1998;Feigelman 1997), or have found a better adjustment in adoptees for educational achievement, alcohol consumption, and affiliation, to give a few examples (Lieberman and Morris 2004;Smyer et al. 1998). ...

Long term effects of adoption: An empirical study of adult adoptees

The Internet Journal of Academic Physician Assistants

... The rule of thumb for assessing VIF value should be less than 10.00 while the threshold for bivariate correlation should not be more than 0.800 (Morris and Lieberman, 2015;Kalnins and Praitis Hill, 2025). The results revealed that the values are lower than the cutoff values and there is no serious multicollinearity issue in the data (see Figure 2). ...

Prediction, Explanation, Multicollinearity, and Validity Concentration in Multiple Regression

... In this study, a crucial question we aimed to address is which features in resting-state EEG drive age prediction. Though not detrimental to the overall prediction accuracy (Mundfrom et al., 2018;Morris and Lieberman, 2018), including highly correlated features in a regression model -especially features extracted from neighboring electrodes or ROIs -would negatively impact the ability to accurately interpret the statistical importance of each independent variable (feature) (Farrar and Glauber, 1967;Kim, 2019). Here, we proposed a viable approach to circumvent this multicollinearity problem by evaluating feature importance at a cluster level using hierarchical clustering and Shapley Additive Explanations (SHAP) values (Fig. 3A). ...

Multicollinearity's Effect on Regression Prediction Accuracy with Real Data Structures

General Linear Model Journal

... With globalisation, humanity has entered a new era and individuals have faced brand new practices and regulations in the new world order that has become a global village (Barber, 2000;Kirkwood-Tucker, Morris, & Lieberman, 2010). As a result of increasing inequalities and changing balances with globalisation, countries have started to experience overpopulation, international migration, environmental degradation and so on, and with this, the relationships and socio-economic environments of individuals in their daily lives have also started to change (Drucker 1993;Kennedy 1993). ...

What Kind of Teachers Will Teach our Children? The Worldmindedness of Undergraduate Elementary and Secondary Social Studies Teacher Candidates at Five Florida Public Universities

International Journal of Development Education and Global Learning