Does statistics anxiety impact academic dishonesty? Academic challenges in the age of distance learning

This study discusses the mediating role of statistics anxiety and motivation in the relationship comprising academic dishonesty, personality traits, and previous academic achievements in three different learning environments (Face to Face -F2F, Planned Online Environment – POE, and Emergency Remote Teaching – ERT). Self-determination theory (SDT) provides a broad psychological framework for these phenomena. Data were collected from 649 bachelor-degree students in the Social Sciences in five Israeli academic institutions. Structural equation modelling was employed to investigate the research variables’ relationships. Findings indicate that statistics anxiety mediates the relationship between personality traits and academic dishonesty in the POE and the ERT learning environments. Findings also indicate the relationship between students’ achievements and academic dishonesty, but only in the ERT learning environment. In contrast, motivation mediates the relationship between students’ achievements and statistics anxiety only in the POE learning environment. This study unveils that learning environments determine the mediating role of statistical anxiety. In digital learning environments (POE, ERT), mediation has been found between students’ personality traits and academic dishonesty. No similar parallel mediation could be established in the physical learning environment, F2F. Thus, we conclude that online courses should be designed according to student-centred approaches.

Page 2 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 both moral and practical implications (Bacon et al. 2020), as students' ethical behaviour transfers over into the job force (Walsh et al. 2021). Thus, professional education shall also focus on the ethical formation (Guerrero-Dib et al. 2020). Recent research (Etgar et al. 2019) has revealed the pivotal role of motivation in students' disposition to AD. According to Self-determination theory (SDT) by Deci (2008, 2020), motivation results from either internal or external incentives, which indicates the domain of a self-initiated activity prompted by some external factor (Locquiao and Ives 2020). SDT provides a broad psychological framework for understanding motivation for AD (Krou et al. 2021).
In addition, knowledge of statistics has been recognized as mandatory in academic education (Trassi et al. 2022) thus; current academic training includes compulsory introductory statistics courses. Occasionally, some students associate these with high anxiety levels (O'Bryant et al. 2021). For some students experiencing Statistics anxiety (SA), this assignment has a negative impact on their academic experiences (Trassi et al. 2022). Previous research on undergraduate social sciences students (Steinberger et al. 2021) unveiled that students' anxiety toward statistics negatively influences learning and academic performance. Moreover, anxiety and inappropriate academic behaviours are related (Zhang et al. 2020). In addition, (Tindall et al. 2021) found that negative emotions influence students' propensity to engage in unethical conduct like plagiarism.
Likewise, research has shown a significant interrelation among attitudes toward statistics, anxiety, and performance, which are determined by students' prior statistics or mathematics education (Peiró-Signes et al. 2021). Scholarly review literature (Cui et al. 2019;Chiang et al. 2022) has indicated that dispositional character and person-related circumstances determine statistics anxiety (SA). Furthermore, research dealing with the influence of SA on student academic performance is vast (O'Bryant et al. 2021), including factors which predict AD (Roe 2022).
Yet, studies on SA, AD and pandemic circumstances are scanty (Steinberger et al. 2021). Our research fills this gap by examining the relationship comprising: AD, SA, personality traits, and motivation in undergraduate students in the social sciences taking an Introduction to Statistics compulsory course in different learning environments (Face to Face -F2F, Planned Online Learning -POE and Emergency Remote Teaching -ERT). Understanding AD's profile and likelihood is key to personalising academic interventions meant to discourage and reduce it and SA manifesting in different learning environments. Furthermore, our research enlightens the mediating roles SA and motivation play in the relationship comprising personality traits, previous achievements, and AD. Thus, the main research question is: To what extent does the relationship among SA, personality traits, previous achievements, and motivation affect AD in the different learning environments (F2F, POE, ERT)?

Academic dishonesty
Effective learning and teaching are key research topics in higher education (Steinberger et al. 2021). Academic integrity is a desideratum for quality education (Ozoliņa and Bēriņa 2021). Quality education is unattainable without respect for academic integrity (Artiukhov and Liuta 2017), and without maintaining quality educational process standards in the educational (Kudeikina et al. 2022). Understanding this has contributed Page 3 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 to expanding scholarly knowledge on academic integrity and preventing AD (Parnther 2020). Whereas academic integrity refers to trustful, respectful, fair, and responsible behaviours (Sefcik et al. 2020), AD refers to offences that include: cheating, plagiarism, fabrication, and facilitation (Etgar et al. 2019). Studies have shown the omnipresence of AD as a normalized student behaviour (Krou et al. 2021;Chiang et al. 2022;Christensen Hughes and Eaton 2022) and that most students engage in AD at some point in their studies (Peled et al. 2019). Furthermore, other studies have indicated that AD is more likely to prevail among certain populations (Hensley et al. 2013). For example, business students have a greater propensity to cheat than non-business students (McCabe et al. 2006).

Statistics anxiety
Statistical literacy has become an essential skill for higher education students for their academic and future professional practice (Berndt et al. 2021), including in business programs (Vaziri et al. 2022). Yet, studies have revealed that students experience problems with learning, understanding, and using basic statistical notions. Statistics anxiety (SA) is one of the most common phenomena following this (Murtonen 2015). SA refers to a negative emotional state or attitude provoked by any form of contact with statistically related content (O'Bryant et al. 2021). Hence, it often interferes with teaching-learning quantitative material. According to a research literature review (Cui et al. 2019), SA's antecedents are: (a) Dispositional factors (personality traits), (b) personal factors (previous academic achievement, motivation) and (c) situational factors (attitudes connected to statistics). The SA six-factor model is a largely acknowledged approach (Levpušček and Cukon 2020), identifying six elements informing SA, which is integrated into the common SA rating scale (STARS) by Cruise et al. (1985). The above encompasses anxious feelings and learners' attitudes towards statistics: Interpretation anxiety -the anxiety following the need to interpret different statistical data; Test and class anxiety -the anxiety manifesting while attending statistics courses and taking statistics tests; Fear of asking for help -the anxiety manifesting while requesting assistance to understand statistics; Computational self-concept -an individual's perception of his mathematical abilities for learning statistics; Worth of statistics -the significance and relevance of learning statistics, and Fear of statistics teachers -students' perceptions of statistics teachers.

Personality traits
The Five-Factor Model (FFM) by McCrae and Costa (1987) is an acknowledged psychological tool for theoretically evaluating and measuring personality traits (Dimitriadis et al. 2017). The FFM was neither designed to identify nor measure ethical conduct (Sleep et al. 2021). Nonetheless, research has shown that traits are crucial for understanding students' disposition to engage in AD (Peled et al. 2019). The FFM divides personality into five different traits: Openness to experience, which expresses love for art, adventure, atypical ideas, and imagination; Conscientiousness, which refers to the tendency to exhibit self-discipline and act dutifully; Extraversion, which is intimately related to engagement with the outer world and often characterizes individuals, who are perceived as fully energetic; Agreeableness, which is associated with the value of getting along with others, individuals possessing this latter trait are often considerate, kind, generous,

Academic dishonesty, statistics anxiety, and personality traits
Research on AD has repeatedly employed the FFM (Eshet et al. 2014). It has been revealed that personality determines cheating behaviour due to its impact on personal beliefs, one's attitude towards learning and studying, and goal achievement approach (Malesky et al. 2022). For example, negative affect predicts plagiarism (Tindall et al. 2021). Furthermore, personality traits are also associated with SA (Chew and Dillon 2014). Previous research on the relationship between personality traits, AD and SA, has suggested that: Students scoring high on openness to experience, who are interested in learning and curious, tend to disapprove of AD (Lee et al. 2020) and exhibit lower anxiety levels (Steinberger et al. 2021). Students scoring high in conscientiousness, with a high propensity to follow the rules, exhibit a low cheating propensity (Giluk and Postlethwaite 2015) and are unrelated to SA (Chew and Dillon 2014). Students scoring high on extraversion are often assertive and prone to cheat (Malesky et al. 2022) and are also positively associated with some components of SA (Agbaria and Mokh 2021). Students scoring high on agreeableness have a significantly negative correlation to AD (Malesky et al. 2022) and SA (Cui et al. 2019). Students scoring high on neuroticism (those scoring low on emotional stability) positively correlate with AD (Muntada 2013) and SA. Thus, we posit: H1: Statistics Anxiety will mediate the relationship between Students' Personality Traits and Academic Dishonesty

Motivational orientation
Motivation can psychologically strengthen and stimulate students' learning processes and activities (Becerra and Almendra 2020). Accordingly, it predicts academic performance (Tonguç and Ozaydın Ozkara 2020;Zalts et al. 2021) as it explains one's intentional behaviours (Shi et al. 2021). Furthermore, motivation is a substantial factor in conditioning anxiety (Luo et al. 2020). For example, students having low mathematical proficiency will display higher anxiety rates (Faber and Drexler 2019) and negative attitudes towards statistics (Bromage et al. 2021). According to Self-Determination Theory (SDT) by Deci and Ryan (2008), motivation can either be intrinsic or extrinsic. Intrinsic motivation refers to the willingness to engage in educational activities based on inherent characteristics (genuine interest and enjoyment). Conversely, extrinsic motivation points to one's incentives for doing something due to external outcomes or rewards. Intrinsic motivation is positively associated with academic success, performance, and self-confidence (Foutz et al. 2021). Studies have pointed out that intrinsic motivation positively Page 5 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 impacts self-confidence and responsibility, while extrinsic motivation relates to incompatible behaviours such as anxiety and indifference towards responsibility (Lavasani et al. 2014). Students scoring high on extrinsic motivation are driven by grades, class rank, and earnings (Zalts et al. 2021). Furthermore, motivation and FFM are positively related to academic performance. Research suggests that according to the different personality traits, there are different motivational orientations (Arniatika 2020). For instance, consciousness and openness to experiences correlate with intrinsic motivation; neuroticism correlates with extrinsic motivation (Müller et al. 2006). Thus, we posit: H2 Students' Motivation will mediate the relationship between Students' Personality Traits and Statistics Anxiety.

Previous achievement, statistics anxiety, and academic dishonesty
Previous academic achievement predicts future academic outcomes (Hensley et al. 2013) and success in statistics courses (Sorge and Schau 2002). A myriad of research has explored the cognitive and affective factors that influence students' performance in statistics. According to widespread conceptions, poor achievements is strongly connected to academic misconduct (Koscielniak and Bojanowska 2019). For example, prior research has found that previous academic performance and SA strongly correlate (Siew et al. 2019;Steinberger et al. 2021). For example, some students experience SA due to their lack of mathematical knowledge (Onwuegbuzie and Wilson 2003). Therefore, linking anxiety to performance leads students to higher procrastination rates and avoid statistics-related tasks (Onwuegbuzie and Wilson 2003). This leads to avoidance behaviour (Hong, Tsai, & Tai, 2021) as a form of compensations strategy to better results (Koscielniak and Bojanowska 2019), inducing AD. Other research has emphasized that ethical value behaviour can be ascribed to motivation, learning strategies, and students' previous achievements (Koscielniak and Bojanowska 2019), including AD. Thus, we posit: H3 Statistics Anxiety will mediate the relationship between Students' Previous Achievements and Academic Dishonesty.

Academic dishonesty since the Covid-19 in different learning environments
Practices of AD before the COVID-19 pandemic have been diverse (Gamage et al. 2020). Scholarly studies before the latter's outburst could determine that technology's proliferation has changed the nature of AD offences (Meiring 2019). Yet, knowledge of the impact of these since Covid-19 is still scanty. Online instruction has grown exponentially since the pandemics' outburst, thus altering educational practices' nature and delivery for years to come (Li and Lalani 2020). Furthermore, responses to the pandemic strengthen the active dimension of education (Crawford 2020), e.g., active learning, independent and critical thinking, individual exploration and participatory development (Armellini et al. 2021). Students adapt to different learning environments in this context, often incompatible with academic attendance obligations (Butler-Henderson and Crawford 2020). Scholars have added that the diversity of educational situations and the implementation of heterogeneous learning technologies have led to different educational theories about how technology impacts education (Venn et al. 2020). Hence, the importance of meticulously clarifying how digital integrity has become a crucial 21st-century skill impacting students' ability to evaluate, handle, and share knowledge ethically and successfully (Miller 2019). Notably, there is a broad scholarly consensus on the pedagogical divergences between online and face-to-face teaching. The mere application of traditional educational approaches to online settings has been shown to be ineffective due to the tensions it often creates (Badiozaman 2021). Contextualizing the above phenomenon is mandatory to understand it fully. The learning environments include psychological, pedagogical, and social features influencing students' achievement (Helms 2014). The last decades have witnessed a transformation in the learning environments. These changes further enabled the implementation of innovative pedagogical approaches entwined with modern technology (Valtonen et al. 2021), like the integration of information & communication technology (ICT) teacher or student-centred approaches (Mesny et al. 2021). Educators frame and decide on the course's structure, content, and educational process (Greenberg et al. 2007). Furthermore, technological advancement has led to digitalized -entirely online (planned online environment-POE) or hybrid modules of traditional face to face (F2F) education. Consequently, this has raised concerns about AD and its different ways of cheating (Ikram and Rabbani 2021). In addition, the Covid-19 pandemic impelled education to transform into online delivery (Turnbull et al. 2021), leading to unplanned online teaching and learning formats (Lowenthal et al. 2020), coined as emergency remote teaching -ERT (Hodges et al. 2020). This has further led education to new challenges (Whalen 2020). Previous scholarly research has shown that students' learning experience and performance differ according to learning environments (Mørk et al. 2020;Maqableh and Alia 2021), including course enrolment and delivery methods: F2F, POE, and ERT. For instance, instructive intensity, disorganization, and oppression likely raise SA levels. Additionally, the learning environment and teachers' interactions influence students' motivation (du Rocher 2020). Studies focusing on statistics learning have compared results obtained from POE and F2F statistics courses. POE instruction is less effective than F2F, as it allows learners to be more concretely exposed to their educator's attitudes and concerns. As a result, performance in POE settings is lower than in F2F ones (Cui et al. 2019). Thus, we posit: H4 There will be differences between learning environments in the relationship comprising Statistics Anxiety, Personality Traits, Motivation, Academic Dishonesty, and Previous Academic Achievement.

Research model
Based on the literature above, the research model presents AD as assumed to be influenced by personality traits and students' previous achievements with the mediation of motivation and SA (Fig. 1).
The research model presents personality traits (measured by extraversion, agreeableness, conscientiousness, openness to experiences, and emotional stability), students' previous achievements (measured by mathematics level, grade point average, and matriculation grade in mathematics) with the mediation of the latent variable of motivation (measured by external regulation, introjected regulations, identifies regulation and intrinsic motivation), and SA (measured by worth of statistics, interpretation anxiety, Page 7 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 test and class anxiety, computational self-concept, fear of asking for help, and fear of statistics teachers) as the factors assumed to influence AD..

Participants and procedure
Data were collected from five Israeli academic institutions students studying for bachelor's degrees in the Social Sciences enrolled in introductory Statistics courses. There was a total of 649 participants, 7% were male and 93% female students. Mean age of participants was 23.5, ranging between 18 and 42, SD 7 years. Questionnaires were administered to the participants in three different course enrolment modalities through an online platform following the approval of the Ethics Committee. More than half of the students (59%) enrolled in POE, 18% in F2F, and 23% in ERT courses. The average time for filling out the questionnaires was 12 min. Fourteen per cent of the participants were excluded from the analysis as their survey instruments were incomplete (less than 80%) or carelessly completed. Among the participants, 6.5% reported high SA (the mean higher than 4 on a scale from 1 to 5

Dependent variables
Academic Dishonesty was measured directly through the Academic Misconduct Scale (Bolin 2004) and indirectly through the Academic Integrity Inventory (Kisamore et al. 2007). and validated these instruments to the Israeli context. The Academic Misconduct Scale comprises 10 items on a five-point Likert scale, in which 1 means "Never" and 5 "Many times". Its reliability is excellent (0.91 Cronbach's alpha). The Academic

Fig. 1 Structural Model for Determinants of Academic Dishonesty in Statistics Courses
Page 8 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 Integrity Inventory consists of 8 items on a five-point Likert scale, in which 1 means "Very unlikely" and 5 "Very likely". Its reliability is acceptable (0.75 Cronbach's alpha).

Mediating variables
Statistics Anxiety -This research uses the Hebrew version of the Statistics Anxiety Rating Scale (H-STARS), which is an abridged version of the STARS scale developed by Cruise et al. (1985). The H-STARS has been adapted to the Israeli context and found reliable and valid (Steinberger 2020). The Hebrew version of STARS comprises 30 items and employs six different subscales: worth of statistics; interpretation anxiety; test and class anxiety; computational self-concept; fear of asking for help; fear of statistics instructors. Participants answer questions about possible anxiety-inducing situations and their attitudes to statistics on a 5-point scale, in which 1 means no anxiety and 5 a great deal thereof. Steinberger (2020) has reported good internal consistency reliability (0.80-0.94). These are consistent with those presented previously in Cruise et al. (1985). Following the authors' recommendation, calculating the overall score averages all questionnaire items, so the higher the score, the higher the anxiety level.
Motivational orientation -We employed the Academic Self-Regulation Questionnaire (SRQ-A) (Ryan and Connell 1989), which evaluates four types of motivation: intrinsic motivation, identified, introjected, and external regulation. Participants answered 17 questions employing a five-point Likert scale, in which 1 means "Not true at all and 5 "Very true". As measured by Cronbach's alpha, the questionnaire's reliability is acceptable (0.75)..

Independent variables
Personality traits -The survey employs the Ten Item Personality Inventory (TIPI) scale by Gosling et al. (2003), which is comprised of 10 items developed to evaluate the personality traits of the participants on a five-point Likert scale, in which 1 means "Not true at all and 5 "Very true". Two statements inform each trait. The reliability of this questionnaire, as measured by Cronbach's alpha is questionable (0.63).. Previous academic achievements are measured according to students' high school mathematics level, grade point average, matriculation grade in mathematics, and course enrolment type..

Plan of analysis
We have analysed the data through Structural Equation Modelling (SEM). Full information maximum likelihood estimates were computed using the Analysis of Moment Structures (AMOS) program (Arbuckle and Wothke 1999). The model was examined for the goodness of fit using χ2, comparative fit index (CFI), and root mean square error of approximation (RMSEA) fit indices. CFI values above 0.90 and 0.95 indicate adequate and good model fit, respectively, and RMSEA values below 0.08 and 0.05 indicate adequate and good model fit, respectively (Browne and Cudeck 1992;Hu and Bentler 1999). In addition, we used descriptive statistics and Pearson Correlations to analyse the data. Reliability analysis was done as well.

Results
The descriptive statistics and correlations between the research variables are presented in Table 1.
The results show significant negative correlations between all the five personality traits and AD. Furthermore, there is a significant negative correlation between identified regulation and AD and significant positive correlations between AD, external and introjected regulation, and intrinsic motivation. There are positive correlations between each of the components of SA and AD. Table 2 presents inter-correlations between statistics anxiety, motivation, personality traits, and the dependent variables of academic misconduct and academic integrity in the three learning environments..
Results show significant negative correlations between the personality traits of openness to experience and emotional stability and statistics anxiety in all three learning environments, as well as between the personality traits of agreeableness and conscientiousness and statistics anxiety, albeit in the ERT sample only. In addition, there is a significant positive correlation between external regulation and statistics anxiety in the POE and ERT learning environments and significant negative correlations between identified regulation and intrinsic motivation and statistics anxiety in all three learning environments. Nonetheless, significant positive correlations between statistics anxiety and dependent variables were found in the POE and ERT samples, though not in the F2F one. The AD variable was modelled by the variables of academic misconduct and academic integrity, by the latent variable of personality, and those of motivation, and of students' previous achievements with the mediation of the latent variable of SA. The data fit the academic dishonesty model marginally well (χ2 = 1,426.37, N = 649, df = 564, p < 0.001, CFI = 0.801, RMSEA = 0.049)..

Academic dishonesty analysis -POE sample
The structural model of academic dishonesty in the POE sample is illustrated in Fig. 2.
The results of the analysis indicate that the variance in AD is explained by students' personality traits with the mediation of SA. Accordingly, the POE sample supports H 2 . SA is the variable having a greater impact on academic misconduct with a total effect of 67%. As shown in Fig. 2, test and class anxiety are among the most influential components of SA. It has one of the highest effects (b = 0.79, p < 0.001), meaning that the higher a student's level of SA as to test and class anxiety, the higher their propensity to cheat. The same applies to the component of SA regarding computational self-concept. It has been found to have a strong significant effect as well (b = 0.77, p < 0.001), while the higher the student's [level of SA regarding] computational self-concept, the higher the probability that they engage in academic misconduct. Interpretation anxiety (b = 0.76, p < 0.001) is a further strong factor influencing academic misconduct. Accordingly, SA regarding interpretation anxiety increases academic misconduct. In addition, personality traits were found to have a significant negative impact on SA (b= -0.34, p < 0.05). All of the five personality traits have a significant effect on the mediating variable of SA: Extraversion (b = 0.16, p < 0.05), agreeableness (b = 0.48, p < 0.05), conscientiousness (b = 0.46, p < 0.001), openness to experience (b = 0.23, p < 0.01) and emotional stability (b = 0.73, p < 0.001). Accordingly, the higher levels of a student's personality traits, the less anxious they is. Motivation was also found to have a negative significant impact on SA (b= -0.36, p < 0.01), while all motivation types have a significant effect on the mediating variable:  Page 11 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 external regulation (b = 0.21, p < 0.001), introjected regulation (b = 0.70, p < 0.001), identified regulation (b = 0.76 p < 0.001) and intrinsic motivation (b = 0.77, p < 0.001). In other words, the higher the student's motivation, the lower they level of SA is.

Academic dishonesty analysis -F2F sample
The structural model of AD in the F2F sample is illustrated in Figure 3. The analysis's results indicate that the variance in AD is explained by students' personality traits and motivation, with no significant effect of SA as a mediator. Therefore no support for the four hypotheses was obtained in the F2F sample. As shown in Fig. 3, personality traits were found to have a negative significant impact on SA (b= -0.55, p < 0.01), while three personality traits have a significant effect on the mediating variable: conscientiousness (b = 0.35, p < 0.05), openness to experience (b = 0.72, p < 0.001) and emotional stability (b = 0.67, p < 0.001). This means that the higher a student's personality traits, the lower they level of SA is. Motivation was also found to have a negatively marginal significant impact on SA (b= -0.36, p = 0.065), while all motivation types have a significant effect on the mediating variable: external regulation (b = 0.58, p < 0.001), introjected regulation (b = 0.98, p < 0.001), identified regulation (b = 0.36 p < 0.001) and intrinsic motivation (b = 0.34, p < 0.01). In other words, the higher a student's motivation, the less anxious they is. Grade point average (b = 0.46, p < 0.05) is a further variable having a significant negative effect on SA. The higher a student's grade point average, the lower they SA is. Gender and age were also found to have a significant effect on SA (b = 0.22, p < 0.05 and b = 0.23, p < 0.01, respectively). Accordingly, women experience greater SA than their male counterparts, and the older the students age, the greater the SA.

Academic dishonesty analysis -ERT sample
The structural model of AD in the ERT sample is illustrated in Fig. 4. The analysis's results indicate that the variance in AD is explained by students' personality traits and students' previous achievement, with the mediation of SA. Therefore, H 2 and H 4 were confirmed in the ERT sample. SA is the variable having the greatest impact on academic misconduct, with a total effect of 49%. As shown in Fig. 4, test and class anxiety are among the most influential components of SA; it has one of the higher effects (b = 0.83, p < 0.001). The higher a student's level of SA as to test and class anxiety, the higher their propensity to cheat. Similarly, the component of SA regarding fear of asking for help was also found to have a strong significant effect (b = 0.82, p < 0.001). The higher a student's level of SA as to fear of asking for help, the higher the probability they engage in academic misconduct. In addition, interpretation anxiety (b = 0.80, p < 0.001) is a further strong factor influencing academic misconduct. Moreover, personality traits were found to have a negative significant impact on SA (b= -0.55, p < 0.001), along with a positive significant effect on motivation (b = 0.46, p < 0.01), while all the personality traits have significant effect on the mediating variables: extraversion (b = 0.36, p < 0.001), agreeableness (b = 0.38, p < 0.001), conscientiousness (b = 0.58, p < 0.001), openness to Notes: β = standardized regression weight; SE, standardized error; CR, critical ratio. *p < 0.05; **p < 0.01, ***p < 0.001 The confidence interval of 95% in Brackets 1 We conducted a Multicollinearity analysis, which showed that there is no multicollinearity between the research variables: VIF of all the research variables ranged from 1.048 to 2.730 (less than 5) and Tolerance ranged from 0.366 to 0.954 (above 0.2) Page 14 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 experience (b = 0.55, p < 0.001) and emotional stability (b = 0.54, p < 0.001). This means that the higher a student's levels of one of the above personality traits, the more motivated and less anxious they is. Another set of variables having a negative significant effect on SA are those related to previous student achievements (b= -0.46, p < 0.05): grade point average (b = 0.53, p < 0.05), mathematics level (b = 0.25, p < 0.05) and matriculation grade in mathematics (b = 0.42, p < 0.05). The higher one's previous student achievements are, the lower they level of SA. Table 3 summarizes the testing results for the research hypotheses. As shown in Table 3, the analysis results indicate that there was no significant indirect effect between personality traits and SA through the mediation of motivation in any of the learning environments. Accordingly, no support for H 1 was obtained. A significant indirect effect between personality traits and AD mediated by SA was found in the POE and the ERT samples, thus partially confirming H 2 . A significant indirect effect between students' achievements and SA through the mediation of motivation was found only in the POE sample, thus partially confirming H 3 . A significant indirect effect between students' achievements and AD through the mediation of SA was found only in the ERT sample, thus partially confirming H 4 ..
As shown in Table 4, the results of the multi-group analysis indicate that there is a significant difference between all course types: POE, F2F, and ERT, thus confirming H 4 .

Discussion
Academic institutions' promotion and maintenance of academic integrity are significant concerns (Chugh et al. 2021). The same applies to digital integrity, which has become a crucial 21st-century skill impacting students' ability to handle and share knowledge ethically (Miller 2019) while maintaining their performance levels (Amigud and Lancaster 2019). In this context, the present research presents for the first time a comparison between academic ethical behaviour, SA, personality traits, and motivation in different learning environments (F2F, POE & ERT). In addition, it relies on Self-Determination Theory and expands the existing literature on students' dishonest behaviour (lack of academic integrity) and their motivations for engaging in this in statistics introductory courses. In line with the scholarly literature (Krou et al. 2021), we believe that understanding the motivational and anxiety-related mechanisms involved in unethical academic behaviours is key to designing future teaching, learning, and assessment approaches (Etgar et al. 2019;Steinberger et al. 2021).
The results show that learning environments (F2F, POE & ERT) affect and play a significant role in interacting with SA, motivation, personality traits, and AD (H 4 ). Moreover, findings show that AD is more prevalent in POE than in F2F and ERT environments. This study's findings improve the model employed in previous studies (Peled et al. 2019;Steinberger et al. 2021) by revealing that SA plays a significant role alongside the circumstances comprising learning environments mediating between personality traits and AD. More concretely, this study shows that learning environments determine the mediating role of SA. In digital learning environments (POE, ERT), mediation has been found between students' personality traits and AD. No similar parallel mediation could be established in the physical learning environment, F2F. In line with the scholarly literature (Whittle et al. 2020), these differences may be attributed to the impacts an instructor's presence has in different learning environments. This difference may be due to the virtual communication and lack of physical presence of academic instructors in both POE and ERT learning modalities, which may increase students' anxiety. However, F2F learning is mostly characterized by a student's direct and immediate interaction with the instructor and fellow students. The lack of a teacher's physical presence may lead to uncertainty and anxiety and directly impact students' ethical disinhibition. Additionally, the differences from examining the two digital environments show that in the ERT one, SA mediates between students' previous achievements and AD. No similar parallel mediation could be found in the POE environment. This difference may be attributed to educational delivery methods (asynchronous vs. synchronous), which affect learning quality and process (Steinberger et al. 2021). Furthermore, the immediate necessity to move to digital learning without prior preparation during the global pandemic has led students to severe distress. These have been compelled to deal with existential health anxiety and a state of ongoing uncertainty while continuing to take a demanding course, potentially awakening SA. In addition, the quality of distance teaching is lower in ERT due to being imposed at once without any prior pedagogical preparation (Hollweck and Doucet 2020). Accordingly, students facing exceptional and extreme situations like this may rely exclusively on their previous academic experience or achievements in studies in general and, more concretely, in mathematics. A successful student may be less anxious about statistics, thus refraining from unethical behaviour. On the other hand, in asynchronous online courses, one's experience of previous academic success is not related to AD as mediated by SA.

Conclusion
Learning is a socially determined activity (Goodhart 2020), as individuals learn from and with others, even at a distance. Hence, online courses should be designed according to student-centred approaches (Rapanta et al. 2020). The foregoing may include: Instructor's immediacy, improved communication, pre-planned real-life based on learning tasks (Neumann et al. 2013), and monitoring of student progress, for which using continuous formative assessment is key (Torres Martín et al. 2021). Such an approach may create an optimal class climate, overcome the limitations of digital learning, and decrease SA and AD. For, instructors' immediacy and direct communication are concrete in F2F instruction, POE and ERT rely on electronically mediated communication. This, in turn, promotes students' sense of self-competence and autonomy throughout their learning processes, thus reducing dishonesty (Kanat-Maymon et al. 2015). Additionally, positive attitudes towards learning statistics are crucial to motivate students and awaken their interest in the subject. These significantly impact the general class climate and student academic performance (Bromage et al. 2021). In this context, recent scholarship has revealed an increasing trend among students to pay external agents to prepare their academic assignments (contract cheating) (Birks et al. 2020). The major Page 16 of 20 Eshet et al. International Journal for Educational Integrity (2022) 18:23 causes for this are students' dissatisfaction with teaching and learning environments, the stress of time, and the perception of cheating opportunities resulting from the current variety of technological possibilities facilitating non-ethical behaviour (Amzalag et al. 2021). Scholars have consequently stressed that deepening student engagement and learning requires that part-time faculty take part in discussing and communicating ideas and creating clear policies and shared tasks (Artiukhov and Liuta 2017).

Limitations and future research
Data were collected before academic institutions formulated clear examination policies to transition to distance learning. Hence, respondents experienced ambiguity regarding the course's evaluation method (test or paper) and could not design unethical behaviour strategies. Nor could they know whether the latter would take place on campuses or be carried out remotely electronically. Like any other empirical model, the present model is a specific theoretical construct analysing and reflecting a given practice (its data). In other words, our model offers a particularized theoretical perspective of a general sociocultural phenomenon. This entails that research, theory, and practice could all potentially benefit from similar tests focusing on additional contexts and employing other predictors. Future research may investigate the impact of hybrid learning environments on the relationship comprising SA and AD.
Abbreviations AD Academic Dishonesty. ERT Emergency Remote Teaching.

F2F
Face To Face.

FFM
Five-Factor Model.

POE
Planned Online Environment.

SA
Statistics Anxiety.