Statistical modelling of predictors influencing instructional technology uptake among teaching professionals

Introduction

The challenge of predicting instructional technology uptake is not new, but it persists because the predictor landscape is crowded and context-dependent [1] . Dozens of variables -from personality traits and pedagogical beliefs to institutional budgets and peer norms -have been proposed as drivers of adoption, yet few research eforts have atempted to rank these variables within a single statistical model that accounts for their shared variance [2] . Without such ranking, educational planners face a practical dilemma: limited resources must be allocated to intervention strategies, but which predictor deserves the largest investment? If perceived usefulness is the dominant driver, then demonstration workshops that showcase too benefits should take priority. If self-eficacy maters most, then hands-on skill-building sessions are the right call. And if institutional support is the botleneck, then no amount of individuallevel training will move the needle unless infrastructure and policy gaps are addressed first [3] . Multiple regression analysis provides a natural framework for addressing this question because it allows simultaneous estimation of each predictor's unique contribution while controlling for the others [4] . Hierarchical regression adds a layer of theoretical discipline by specifying the order in which predictors enter the model, testing whether later variables add explanatory power beyond what earlier ones already account for [5] . Together, these techniques yield a predictor hierarchy -a ranked list of variables ordered by their incremental contribution to the outcome -that can directly inform resource allocation decisions [4] . The Malaysian higher education context is well suited for this type of modelling exercise. Malaysia has invested heavily in campus-level ICT infrastructure under successive national education blueprints, yet uptake among teaching professionals remains uneven across institutions and disciplines [6] . Prior Malaysian research has documented high rates of personal device ownership and internet access among faculty, suggesting that infrastructure availability alone cannot explain the variation in technology uptake [7] . Individual-level predictors -beliefs, skills, and social influences -appear to carry considerable weight, but their relative importance has not been formally quantified through regression modelling in the Malaysian context. Six candidate predictors were selected for inclusion in the present model based on their prevalence in the educational technology adoption literature and their theoretical alignment with the Technology Acceptance Model and Social Cognitive Theory [1,8] . Perceived usefulness reflects the belief that technology will improve teaching outcomes. Technology selfefficacy captures confidence in one's ability to operate digital tools. Professional development hours measure formal training exposure. Institutional support indexes the availability of technical assistance and policy backing. Peer influence captures the social dimension of adoption. And years of teaching experience serves as a proxy for career-stage efects documented in comparable contexts [9] . Against this background, the present research aimed to build and compare regression models predicting instructional technology uptake among teaching professionals in Kuala Lumpur, Malaysia, and to establish a predictor hierarchy that institutional decision-makers can use to target their adoptionsupport strategies.

Materials and Methods

Materials

This research was conducted across six higher education institutions in Kuala Lumpur, Malaysia. Institutions were selected to represent a mix of public universities [3] , private universities [2] , and polytechnics [1] , ensuring variation in resource levels and student populations. The target population comprised all full-time teaching professionals with at least one academic year of service, totalling 412 individuals across the six institutions. Through proportionate stratified random sampling -stratifying by institution type and faculty -a sample of 250 was drawn. After excluding incomplete responses, 228 usable questionnaires remained (response rate: 91.2%). Ethical approval was obtained from the National Institute of Mathematical Sciences Research Ethics Panel. All participants signed informed consent forms.

Methods: A 34-item structured questionnaire was used. Section A: collected demographics: gender, age range, years of teaching experience, academic qualification, and discipline. Section B measured six predictor variables: perceived usefulness (7 items), technology self-eficacy (6 items), professional development hours (single continuous item plus 4 perception items), institutional support (5 items), peer influence (4 items), and years of experience (continuous). Section C measured instructional technology uptake (7 items covering frequency, depth, and breadth of technology use). All atitudinal items used a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). Items were adapted from validated scales in the educational technology literature [1,8] .

Pilot research

The instrument was pilot-tested with 26 teaching professionals from a seventh institution excluded from the main sample. Cronbach's alpha values ranged from 0.80 (peer influence) to 0.91 (perceived usefulness). Exploratory factor analysis on the pilot data confirmed the sixfactor structure with eigenvalues exceeding 1.0 and factor loadings above 0.45. One perceived usefulness item was reworded after pilot feedback. Data analysis used SPSS version 28 and R version 4.2 for model comparison. Bivariate Pearson correlations tested associations among all variables. Two regression approaches were employed: (a) enter-method multiple regression with all six predictors entered simultaneously, and (b) stepwise hierarchical regression entering predictors in theoretically determined order -perceived usefulness first, then self-eficacy, professional development, institutional support, peer influence, and experience. Model fit was compared using adjusted R², Akaike Information Criterion (AIC), and Variance Inflation Factors (VIF) for multicollinearity diagnostics. Significance was set at α = 0.05.

Model description and specification

The enter-method regression model took the form: Uptake =: β ₀+β ₁ (PU)+β ₂ (SE)+β ₃ (PD)+ β ₄ (IS)+β ₅ (PI)+β ₆ (Exp)+ε, where PU = perceived usefulness, SE = technology self-eficacy, PD = professional development hours, IS = institutional support, PI = peer influence, Exp = years of experience, and ε = error term. The hierarchical model entered predictors in six sequential steps based on theoretical priority from the Technology Acceptance Model and Social Cognitive Theoryindividual belief constructs first, then environmental and demographic variables. This ordering allowed each step's incremental R² change to be tested for significance, producing the predictor hierarchy that was the primary analytical goal.

Results

The sample of 228 teaching professionals comprised 131 females (57.5%) and 97 males (42.5%), with a mean teaching experience of 10.7 years (SD = 7.2). The overall technology uptake mean was 3.67 (SD = 0.73). Among the six predictors, perceived usefulness recorded the highest mean (3.94, SD = 0.59), while peer influence scored lowest (3.28, SD = 0.78). The correlation heatmap and regression results are presented below.

Comprehensive Interpretation:

In the hierarchical model, Step 1 (perceived usefulness) accounted for 39.7% of uptake variance (ΔR²=0.397, p < 0.001). Step 2 (technology self-eficacy) added 3.8% (ΔR²=0.038, p < 0.001) Step 3 (professional development hours) contributed 1.4% (ΔR²=0.014, p=0.018). Step 4 (institutional support) added 1.2% (ΔR²=0.012, p=0.034). Steps 5 (peer influence, mathsfp=0.104) and 6 (years of experience, mathsfp=0.006 but ΔR²=0.003) did not meaningfully improve model fit beyond Step 4, although experience retained a significant negative coeficient in the full model (β=-0.16). The four-predictor hierarchical model was preferred based on AIC (412.7 vs. 418.3) and parsimony. All VIF values remained below 2.8, and Cook's distance values stayed under 0.04, confirming absence of influential outliers. The predictor hierarchy -usefulness > selfeficacy > training > supportprovides a clear order of intervention priority for institutions seeking to raise technology uptake among their teaching staf.

Discussion

Perceived usefulness emerging as the strongest single predictor (β=0.34, ΔR²=0.397) is consistent with the central tenet of the Technology Acceptance Model and with meta-analytic evidence from Scherer et al. [1] , who reported that usefulness typically accounts for the largest share of adoption variance across educational technology research. The mechanism is straightforward: teaching professionals who believe that a tool will improve their instructional outcomes -whether through more engaging presentations, quicker feedback, or easier content distribution -are more inclined to invest the efort required to learn and use it [8] . Technology self-eficacy ranking second (β=0.27) aligns with Social Cognitive Theory's prediction that confidence in one's ability to perform a behaviour is a proximal determinant of that behaviour [10] . However, the relatively modest incremental contribution (ΔR²=3.8%) compared to usefulness suggests that self-eficacy amplifies uptake among those who already see the tool as useful, rather than independently driving it. This distinction maters for intervention design: self-eficacy training is most productive when paired with, not substituted for, usefulness demonstrations [3] . The negative efect of years of experience (β=-0.16) confirms career-stage paterns observed in comparable contexts [9] . That this variable retained significance in the full model but added litle incremental variance (ΔR²=0.3%) in the hierarchical model indicates that its negative influence is partly mediated through lower self-eficacy and lower perceived usefulness among longer-serving professionals. Direct targeting of experience as an intervention lever is therefore less eficient than addressing the mediating beliefs it operates through. Peer influence failing to reach significance (p=0.104) was somewhat unexpected given the emphasis placed on collegial norms in the Difusion of Innovations literature [11] . One possible explanation is that the Malaysian institutional culture prioritises individual performance evaluation over collaborative teaching practice, weakening the channel through which peer behaviour might influence personal technology decisions [6] .

Practical Applications

The predictor hierarchy translates directly into a staged intervention strategy. Stage 1: demonstrate usefulness through discipline-specific showcases that connect technology features to concrete teaching tasks -this single action targets the predictor responsible for nearly 40% of uptake variance. Stage 2: build self-eficacy through structured, hands-on workshops with graduated dificulty levels, addressing the second-ranked predictor. Stage 3: increase professional development hours by embedding technology training into existing faculty development calendars rather than treating it as optional add-on activity. Stage 4: strengthen institutional support by assigning dedicated technical assistance personnel to high-need departments. This staged approach allocates resources in proportion to each predictor's empirical contribution.

Conclusion

The main contribution of this research lies in providing a statistically grounded predictor hierarchy for instructional technology uptake that can serve as a resource allocation guide for Malaysian higher education institutions. By comparing enter-method and hierarchical regression models across six candidate predictors, the investigation moved beyond the common practice of listing significant correlates and instead quantified the incremental explanatory power each predictor adds to the model. Supporting evidence confirmed several paterns. Perceived usefulness dominated the predictor landscape, accounting for nearly 40% of uptake variance on its own -a share so large that no intervention strategy can aford to overlook it. Technology self-eficacy and professional development hours, while secondary, made statistically significant incremental contributions that justify dedicated training investments. Institutional support reached significance but contributed modestly, suggesting that support infrastructure maters more as a necessary condition than as a primary driver. Peer influence fell short of significance in this sample, challenging assumptions drawn from difusion theory about the power of collegial norms in academic setings. Years of experience exerted a significant negative efect but added negligible incremental variance, indicating that career-stage interventions should target the mediating beliefs (usefulness and self-eficacy) rather than experience itself. The significance of these findings for the broader educational technology field is twofold. First, they demonstrate that predictor importance is not uniform -a handful of variables account for the overwhelming majority of uptake variance, and

References

1. Scherer R, Siddiq F, Tondeur J. The technology acceptance model (TAM): a meta-analytic structural equation modelling approach to explaining teachers' adoption of digital technology in education. Comput Educ. 2019;128:13-35.
2. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS Q. 2003;27(3):425-478.
3. Ertmer PA, Otenbreit-Leftwich AT, Sadik O, Sendurur E, Sendurur P. Teacher beliefs and technology integration practices: a critical relationship. Comput Educ. 2012;59(2):423-435.
4. Field A. Discovering statistics using IBM SPSS statistics. 5th ed. Sage; 2018.
5. Cohen J, Cohen P, West SG, Aiken LS. Applied multiple regression/correlation analysis for the behavioral sciences. 3rd ed. Routledge; 2003.
6. Samuel RJ, Abu Bakar Z. The utilisation and integration of ICT tools in promoting English language teaching and learning. Int J Educ Dev Using ICT. 2006;2(2):4-14.
7. Kakar N, Manjula N, Nadeerpoor MA, Yousafzai I. Teachers' perceptions of factors encouraging the use of information and communication technology (ICT) in agricultural education. South Asian J Agric Sci. 2022;2(2):90-93.
8. Bandura A. Self-eficacy: toward a unifying theory of behavioral change. Psychol Rev. 1977;84(2):191-215.
9. Inan FA, Lowther DL. Factors afecting technology integration in K-12 classrooms: a path model. Educ Technol Res Dev. 2010;58(2):137-154.
10. Compeau DR, Higgins CA. Computer self-eficacy: development of a measure and initial test. MIS Q. 1995;19(2):189-211.
11. Rogers EM. Difusion of innovations. 5th ed. Free Press; 2003.
12. Teo T, Fan X, Du J. Technology acceptance among preservice teachers: does gender mater? Australas J Educ Technol. 2015;31(3):235-251.
13. Hew KF, Brush T. Integrating technology into K-12 teaching and learning: current knowledge gaps and recommendations for future research. Educ Technol Res Dev. 2007;55(3):223-252.
14. Agarwal R, Prasad J. A conceptual and operational definition of personal innovativeness in the domain of information technology. Inf Syst Res. 1998;9(2):204-215.
15. Sánchez RA, Hueros AD. Motivational factors that influence the acceptance of Moodle using TAM. Comput Hum Behav. 2010;26(6):1632-1640.
16. Teo T, Zhou M. The influence of teachers' conceptions of teaching and learning on their technology acceptance. Interact Learn Environ. 2017;25(4):513-527.
17. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. Springer; 2002.