Chapter Six: Phase I Quantitative Analysis of Cross-Sectional Research
6.6 Path Models
6.6.8 Model 4a path analytic model of prediction of AUDIT
consumption. The model indicated close fit to the data (χ2 = 30.81, df = 18, p=
.030; TLI = .971, CFI = .986, RMSEA = .062, PCLOSE =.275). The model predicted 61% of the variance in intention to behave but only 51% of the variance in AUDIT consumption. Intention was a strong predictor of AUDIT consumption (β = .306, p<.001). Attitude was a strong predictor of intention (β = .474, p<.001).
Subjective norm (β = -.122, p<.05), willingness two (β = .247, p<.05) and past behaviour (β = .157, p<.05) were significant predictors of intention. In relation to AUDIT consumption intention (β = .306, p<.001) and past behaviour (β = .248, p<.001) were the strong predictors. Willingness two was a significant predictor of AUDIT consumption (β = .211, p<.05).
Past behaviour was a strong predictor of identified regulation (β = -.304, p<.001) and intrinsic motivation (β = -.263, p<.001). Past behaviour was a
significant predictor of amotivation (β = .169, p<.05), introjected regulation (β = -.216, p<.05), attitude (β = .163, p<.05) and subjective norm (β = -.140, p<.05).
Past behaviour was a strong predictor of willingness (β = .252, p<.001). Past behaviour was a significant predictor of intention (β = .157, p<.05). Past behaviour was a strong predictor of AUDIT consumption (β = .248, p<.001).
Intrinsic motivation was a significant predictor of PWM (β =- .191, p<.05), Amotivation was a significant predictor of attitude (β = .135, p<.05) and PWM (β = .214, p<.05). Introjected regulation was a strong predictor of attitude (β = -.360, p<.001). Identified regulation was a significant predictor of willingness (β =- .155, p<.05). Amotivation was a strong predictor of PBC (β = -.315,
p<.001).
145
Attitude was a significant predictor of willingness two (β = .229, p<.05).
Attitude was a strong predictor of intention (β = .474, p<.001). Subjective norm was a significant predictor of intention (β = -.122, p<.05). Willingness two was a strong predictor of intention (β = .247, p<.001). Intention was a strong predictor of AUDIT consumption (β = .306, p<.001). Willingness was a significant
predictor of AUDIT consumption (β = .211, p<.05) (see Model 4a in Figure 6.11, variance explained and path coefficients for path models in Appendix B Table B13 and B14)
146 Figure 6.11. Path analytic model of predicting AUDIT consumption (***p < .001, ** p<.05)
147
6.6.9 Model 4b path analytic model of predicting AUDIT problems. The model indicated a close fit to the data (χ2 = 43.82, df = 18, p= .001; TLI = .855, CFI = .971, RMSEA = .088, PCLOSE =.32). The model predicted 61% of the variance in intention to behave but only 37% of the variance in AUDIT problems. Interestingly intention was not a significant predictor of AUDIT problems (β=.156, p=.097). Attitude (β = .474, p <.001) and willingness two (β = .247, p <.001) were strong predictors of intention. Subjective norm (β = -.122, p <.05) and past behaviour (β = .157, p <.05) were significant predictors of intention.
In relation to prediction of AUDIT problems PBC was a strong predictor of AUDIT problems (β = -.325, p <.001), willingness (β = .191, p <.05) and past behaviour (β = .217, p <.05) were significant predictors of AUDIT problems.
Past behaviour was a strong predictor of identified regulation (beta β = -.304, p <.001) and intrinsic motivation (β = -.263, p <.001). Past behaviour was a significant predictor of amotivation (β = .169, p <.05) and introjected regulation (β = -.216, p <.05). Past behaviour was also significant predictor of attitude (β = -.117, p <.05). Past behaviour was a significant predictor of subjective norm (β = -.140, p <.05). Past behaviour was a strong predictor of willingness (β = .252, p <.001). Past behaviour was a significant predictor of intention (β = .157, p <.05) and AUDIT problems (β = .217, p <.05).
Regards STD components, intrinsic motivation was a significant predictor of PWM β
= -.191, p <.05). Amotivation was a significant predictor of attitude (beta β = .135, p <.05) and PWM β = .214, p <.05). Introjected regulation was a strong predictor of attitude (β =- .360, p <.001). Identified regulation was a significant predictor of willingness (β = -.155, p
<.05). Amotivation was a strong predictor of PBC (β = -.315, p <.001).
148
Attitude was a significant predictor of willingness (β = .229, p <.05). Attitude was a strong predictor of intention (β = .474, p <.001). Subjective norm was a significant predictor of intention (β = -.122, p <.05). Willingness was a strong predictor of intention (β = .247, p
<.001). PBC was a strong predictor of AUDIT problems (β = -.325, p <.001). Willingness was a significant predictor of AUDIT problems (β = .191, p <.05) (see Model 4b in Figure 6.12, variance explained and path coefficients for path models in Appendix B Table B15 and B16).
149 Figure 6.12. Path analytic model of predicting AUDIT problems
(***p < .001, ** p<.05)
150
6.6.10 Model 4c Path analytic model of predicting AUDIT total. The model indicated close fit to the data (χ2 = 39.79, df = 18, p= .002; TLI = .882, CFI
= .997, RMSEA = .080, PCLOSE =.067). The model predicted 61% of the variance in intention to behave but only 50% of the variance in AUDIT total.
Intention was a significant predictor of AUDIT total (β = .238, p <.05). Also from the table, it can be seen that willingness two (β = .247, p <.001) and attitude (β = .474, p<.001) were strong predictors of intention. Subjective norm (β = -.122, p <.05) and past behaviour (β = .157, p <.05) were also significant predictors of intention.
Past behaviour was a strong predictor of identified regulation (β = -.304, p<.001), intrinsic motivation (β = -.263, p<.001), willingness two (β = .252, p<.001) and AUDIT total (β = .268, p<.001). Past behaviour was a significant predictor of amotivation (β = .169, p <.05), introjected regulation (β = -.216, p
<.05), attitude (β = -.117, p <.05) and subjective norm (β = -.140, p <.05) and intention (β = .157, p <.05).
In relation to SDT components, intrinsic motivation was a significant predictor of PWM (beta β = -.191, p <.05), amotivation was a significant predictor of attitude (β = .135, p <.05) and PWM (β = .214, p <.05), introjected regulation was a strong predictor of attitude (β = -.360, p<.001). Identified regulation was a significant predictor of willingness two (β = -.155, p <.05). Amotivation was a strong predictor of PBC (β = -.315, p<.001). Identified regulation was a significant predictor of AUDIT (β = -.122, p <.05).
151
Attitude was a significant predictor of willingness (β = .229, p <.05), attitude was a strong predictor of intention (β = .474, p<.001). Subjective norm was a significant predictor of intention (β = -.122, p <.05), willingness was a strong predictor of intention (β = .247, p<.001). Intention was a significant predictor of AUDIT total (β = .238, p <.05). PBC was strong predictor of AUDIT (β = -.211, p<.001), willingness was a strong predictor of AUDIT (beta β = .216, p<.001) (see Model 4c in Figure 6.13, variance explained and path coefficients for path models in Appendix B Table B17 and B18).
152 Figure 6.13. Path analytic model of predicting AUDIT total
(***p < .001, ** p<.05)
153
6.6.11 Model 4d path analytic model of predicting binge amount. The model indicated close fit to the data (χ2 = 32.24, df = 18, p= .021; TLI = .921, CFI
= .984, RMSEA = .065, PCLOSE = .226). The model predicted 61% of the variance in intention to behave but only 45% of the variance in frequency of drinking. Intention was a significant predictor of binge amount (β = .240, p<.05).
Also from the table, it can be seen that attitude (β = .474, p <.001) and willingness (β = .247, p<.001) were strong predictors of intention. Subjective norm (β = -.122, p<.05) and past behaviour (β = .157, p<.05) were significant predictors of intention. In relation to binge amount past behaviour (β = .333, p<.001) was a strong predictor of binge amount. Identified regulation (β = -.182, p<.05),
intention (β = -.240, p<.05) and willingness (β = .171, p<.05) two were significant predictors of binge amount.
Past behaviour was a strong predictor of identified regulation (β = -.304, p
<.001) and intrinsic motivation (β = -.263, p <.001). Past behaviour was
significant predictor of amotivation (β = .169, p<.05) and introjected regulation (β
= -.216, p<.05). Past behaviour was a significant predictor of attitude (β = -.163, p<.05). Past behaviour was a significant predictor of subjective norm (β = -.140, p<.05). Past behaviour was a strong predictor of willingness (β = .252, p <.001).
Intrinsic motivation was a significant predictor PWM total (β = -.191, p<.05). Amotivation was a significant predictor of attitude (β = .135, p<.05).
Introjected regulation was a strong predictor of attitude (β = .135, p <.001).
Identified regulation was a significant predictor of willingness (β = -.155, p<.05).
Amotivation was a strong predictor of PBC (β =-.315, p <.001).
154
Attitude was a significant predictor of willingness (β = .229, p<.05).
Attitude was a strong predictor of intention (β = .474, p <.001) (see Model 4d in Figure 6.14, variance explained and path coefficients for path models in Appendix B Table B19 and B20).
155 Figure 6.14. Path analytic model of predicting binge amount
(***p < .001, ** p<.05)
156
6.6.12 Model 4e Path analytic model of predicting frequency. The model indicated close fit to the data (χ2 = 35.52, df = 18, p= .008; TLI = .899, CFI
= .980, RMSEA = .072, PCLOSE =. 139). The model predicted 61% of the variance in intention to behave but only 31% of the variance in frequency of drinking. Intention was a strong predictor of frequency (β = .367, p <.001). Also from the table, it can be seen that attitude (β = .474, p <.001) and willingness two (β = .247, p <.001) were strong predictors of intention. Subjective norm (β = -.122, p<.05) and past behaviour (β = .157, p<.05) were significant predictors of intention. In relation to predicting frequency past behaviour (β = .341, p <.001) and intention (β = .367, p <.001) were only strong predictors of frequency.
Past behaviour was a strong predictor of identified regulation (β = -.304, p
<.001) and intrinsic motivation (β = -.263, p <.001). Past behaviour was a
significant predictor of amotivation (β = .169, p<.05) and introjected regulation (β
= -.216, p<.05). Past behaviour was a significant predictor of attitude (β = -.117, p<.05). Past behaviour was a significant predictor of subjective norm (β = -.252, p<.05). Past behaviour was a strong predictor of willingness (β = .252, p <.001).
Intrinsic motivation was a significant predictor of PWM total (β = -.191, p<.05). Amotivation was a significant predictor of attitude (β = .135, p<.05) and PWM total (β = .214, p<.05). Introjected regulation was a strong predictor of attitude (β = -.360, p <.001). Identified regulation was a significant predictor of willingness (β = -.155, p<.05). Amotivation was a strong predictor of PBC (β =-.315, p <.001).
157
Attitude was a significant predictor of willingness (β = .229, p<.05).
Attitude was a strong predictor of intention (β = .474, p <.001) (see Model 4e in Figure 6.15, variance explained and path coefficients for path models in Appendix B Table B21 and B22).
158 Figure 6.15. Path analytic model of predicting frequency
(***p < .001, ** p<.05)
159
6.6.13 Model 4f Path analytic model of predicting binging. The model indicated close fit to the data (χ2 = 32.23, df = 18, p= .021; TLI = .921, CFI = .984, RMSEA = .065, PCLOSE = .226). The model predicted .61% of the variance in intention to behave but only .45% of the variance in binge recorded.
Intention was a significant predictor of binge recorded (β = .241, p <.05). Also from the table, it can be seen that attitude (β = .474, p <.001) and willingness two (β = .246, p<.001) were strong predictors of intention. Subjective norm (β = -.112, p <.05) and past behaviour (β =.157, p <.05) were also significant predictors of intention (β = -.112, p <.05). In relation to prediction of binging past behaviour (β = .333, p<.001) was a strong predictor of binging and identified regulation (β = -.182, p <.05), intention (β = .241, p <.05), willingness two (β =.171, p <.05) were significant predictors of binging.
In terms of SDT elements, intrinsic motivation was a significant predictor of PWM total (β = -.191, p<.05), amotivation was a significant predictor of attitude (β = .135, p<.05), amotivation was a significant predictor of PWM (β = .214, p<.05), introjected regulaton was a strong predictor of attitude (β = -.360 , p<.001), identified regulation was a significant predictor of willingness two (β = -.155, p<.05), amotivation was a strong predictor of PBC (β = -.315 , p<.001), identified regulation was a significant predictor of binge recorded (β = -.182, p<.05).
Past behaviour was a strong predictor of identified regulation (β = -.304, p<.001), intrinsic motivation (β = -.263, p<.001) and a significant predictor of amotivation (β = .169, p<.05) and introjected regulation (β = -.216, p<.05). In
160
relation to TPB components past behaviour was a significant predictor of attitude (β = .163, p<.05) and subjective norm (β = -.140, p<.05). In relation to PWM components, past behaviour was a strong contributor of willingness two (β = .252, p<.001). Attitude was a significant predictor of willingness (β = .229, p<.05) (see Model 4f in Figure 6.16, variance explained and path coefficients for path models in Appendix B Table B23 and B24).
161 Figure 6.16. Path analytic model of predicting binge recorded
(***p < .001, ** p<.05)
162 6.7 Discussion
The aim of the present study was to test several models, which was an integration of theory of planned behaviour, self-determination theory, prototype willingness model, social learning theory components and several personality variables to increase the predictive properties of the models. For example, Ajzen (2011) suggests theory of planned behaviour could be improved by adding extra personality variables, which would increase its predictive properties.
When theory of planned behaviour was integrated with self-determination theory, as in previous research motivation, autonomous and controlled forms showed to be predicting cognitions: attitudes, subjective norm, perceived behavioural control and self-efficacy
(Caudwell & Hagger, 2015; Hagger et al., 2012). Initial 5 models were hypothesised based on Hagger et al. (2012). In current research it both autonomous forms of motivation and
controlled added to attitude, subjective norm, self-efficacy and PBC. Identified regulation predicted subjective norm, students who think that their drinking behaviour will be approved by significant others have internalised motivation to perform healthy behaviours. It is
consistent with previous research as subjective norm always positively related to health behaviours (de Vries, Dijkstra, & Kuhlman, 1988) and identified regulation, which is
autonomous motivation, was related to healthy behaviours, as it is a type of motivation which is internal and directed in achieving highly valued goal. Hagger et al. (2012) reported
identified regulation to be the most significant variable in predicting keeping alcohol use within guideline limits.
In current research, identified regulation and alcohol behaviour was mediated by subjective norm and contributed to intention, whereas Hagger et al. (2012) reported it to be attitudes, PBC and intention. SDT constructs have been used to explain reflective evaluation
163
towards performing the behaviour and they have been mediating with social cognitive components (Hagger et al. 2012). In addition, in previous research it was argued that subjective norm would not be related to autonomous forms of motivation, as it would be about social factor in other words not internalised motivation but controlled, would have been more related to subjective norm (Caudwell & Hagger, 2015; Hagger et al, 2012). It seems that perceiving that important other approves or disapproves is related to more internalised form of motivation, and makes this belief internalised. Introjected regulation predicted attitude and self-efficacy, external regulation contributed to attitude and PBC. Caudwell and Hagger (2015) found that autonomous form of motivation did not contribute towards PBC, the authors concluded people’s perception of control is not related to autonomous motives.
Cooke and French (2011) found that predictive properties of PBC changes when the
timeframe to perform the behaviour is included. PBC was significant predictor of intention to binge drinking next week but not today or tomorrow (Cooke & French, 2011), authors
suggested to investigate further why PBC would not predict intention on an occasion. In present study external regulation (controlled motivation) contributed to PBC, perceived control negatively linked to motivation to stay within safe limits based on external influence.
It can be interpreted that students who are keeping their alcohol consumption within safe limits to achieve a reward or avoid negative consequences have less control over the
behaviour. Caudwell and Hagger (2015) suggested people seem to engage in pre-drinking for controlled reasons (to avoid guilt, conform, to gain reward or to avoid negative
consequences) to which determines lower perception of control over the behaviour than social approval.
Similar results were reported by Caudwell and Hagger (2015), controlled motivation was negatively related to PBC. None of the SDT components were directly linked to
behaviour which means that motives do not have spontaneous effect on consumption
164
(Hagger, Chatzisarantis & Harris, 2006). Attitude and self-efficacy were related to the intention positively, and to subjective norm negatively. Positive attitude towards the behaviour and the ability to perform the behaviour were predictors of intention to drink.
Students who thought the behaviour was not approved by significant other were less likely to perform the behaviour. Although, Cooke and French (2011) reported that predictive utilities of TPB changes depending whether the data was collected within the context (e.g., a bar) or not (e.g., library). The context effected to subjective norm-intention relationship but not attitude-intention, PBC- intention relationship (Cooke & French, 2011). Similar predictive properties of two models were observed while predicting frequency, units consumed in a single occasion and AUDIT consumption. Models predicted 17%, 21%, 29%, 16% and 35%
variance in outcome variables (Figures 6.4-6.8). Intention was predicted by attitude,
subjective norm and self-efficacy but not PBC. Five models predicted 52%, 51%, 50%, 51%
and 50% variance in intention.
In regards to mediation effect, autonomous motivation (identified regulation) and intention was mediated by subjective norm confirming Caudwell and Hagger’s (2015) results, in addition they found attitude was a second mediator. The mediation effect of identified regulation is in line with Amiot, Sansfacon, and Louis (2013) who identified belief about social influences to be more internalised, which is true to our sample. It is not controlled by social influences and cannot be interpreted as controlled influence.
In relation to the theory of planned behaviour components statistically significant effect was found of attitude, subjective norm and self-efficacy on intention. As it was mentioned before variance of 17%, 21%, 29%, 16% and 35% was predicted of various alcohol related behaviours (see Figure 6.4-6.8). Very small percentage predicted shows models to be inadequate in evaluating the behaviour (Caudwell & Hagger, 2015).
165
The following model (see Figure 6.9) was hypothesised based on Hagger et al. (2012) and Todd et al. (2014) in which identified regulation from SDT, attitude, PBC and subjective norm and intention from theory of planned behaviour, prototype and willingness from PWM and past behaviour was included to predict AUDIT consumption. Identified regulation is negatively related to willingness and attitude, and is different to previous 5 models. In previous 5 models identified regulation did not contribute to attitude at all, though the result is in line with Caudwell and Hagger (2015) and Hagger et al., 2012. Identified regulation was positively related to subjective norm (Caudwell & Hagger, 2015; Hagger et al., 2012). The more person is motivated to stay within safe limits because of highly valued goal, the more he is internally motivated to keep within safe limits, the less willing to drink. Additionally, he has less positive attitude towards drinking. The result is in line with Amiot et al. (2013) and Caudwell and Hagger (2015) who reported relation of autonomous motivation on subjective norm. Past behaviour was positively related to attitude, willingness, intention and AUDIT consumption and negatively to identified regulation and subjective norm. The more people drank in the last 6 month. The more people drank the last 6 months the more positive attitude, willingness, intention they had towards drinking. In addition, past behaviour positively
predicted present alcohol use. Students who consume seem to have less autonomous
motivation (identified regulation) to stay within safe limits. Positive attitude predicted more willingness to drink. In addition, willingness was also predicted by past behaviour and identified regulation, all together 64% variance in willingness have been predicted. In previous research when PWM and TPB components were investigated, Rivis et al. (2011) found 47% of variance in willingness and 65% of variance in drinking and driving behaviour.
Different to present research willingness was predicted with subjective norm, PBC, prototype evaluation and the interaction between prototype evaluation and prototype similarity (Rivis et al., 2011). PBC was not significant predictor in this model. Subjective norm was negatively
166
related to intention. Intention and willingness added to the variance predicted towards AUDIT consumption. Meta analytic review by Todd et al. (2014) reported that willingness improved the predictability of behaviour for 4.9% over and above intention and the results confirm willingness to be a construct to contribute further to TRA/TPB constructs.
Hagger et al. (2012) mentioned when past behaviour is included in the model, it wiped out the relation between psychological variables and behaviour it would invalidate the model and would confirm behaviour being predicted by previous behaviour. The model (Figure 6.6) confirmed that there is still relation between cognitive constructs intention and behaviour, though past behaviour is contributing to most of the variables in the model.
Last model (see Figure 6.10) was hypothesised based Simons et al. (2005). In
Simons et al.’s (2005) study included gender, extraversion, neuroticism, impulsivity, drinking to cope, expectancy, AUDIT consumption and AUDIT problems. Gender was negatively related to AUDIT consumption and extraversion did not contribute towards any variable.
Neuroticism positively related to AUDIT problems and negatively to expectancy. Impulsivity was also positively related to expectancy and AUDIT problems. Expectancy positively contributed to drinking to cope. Drinking to cope positively predicted AUDIT consumption and AUDIT problems.
As autonomous motivation came out significant in the regression analysis during the pilot study it was decided to include it in line with other SDT constructs (see Figure 6.11-6.16). the following 6 models included SDT, TPB, PWM and past behaviour to predict various alcohol behaviours. Interestingly the same patterns have been observed among the following outcome variables: AUDIT consumption, AUDIT problems. Identified regulation did not predict any of TPB components but willingness. Intrinsic motivation predicted to prototype. Amotivation was significant predictor of attitude, PBC and prototype. Introjected regulation was negatively linked to attitudes. From TPB constructs attitude was significant
167
predictor of willingness, subjective norm was negatively predicted intentions. PBC again did not show any significant relation with either intention or outcome variable. Willingness contributed to intention and outcome variable (AUDIT consumption and AUDIT problems).
It shows that spontaniouty of willingness. Past behaviour was adding to the identified
regulation, intrinsic motivation, amotivation, introjected regulation, attitude, subjective norm, willingness intention and outcome variable (AUDIT consumption and AUDIT problems).
Past behaviour was creating noise in previous studies too (Hagger et al., 2012)
The following 4 models were constructed to check for AUDIT total, binge amount, frequency, binge recorded (see Figures 6.13, 6.14, 6.15, 6.16). In addition to the relationship reported between the constructs in the previous 2 models (see Figure 6.11 and 6.12), in the following 4 models (see Figure 6.13, 6.14, 6.15, 6.16) identified regulation had a direct effect on outcome variables’. Previous research highlighted the importance of identified regulation in drinking context (Cooke & French, 2011) but so far none of the literature reported
identified regulation to be directly linked to outcome variables. Table 6.20 describes time 1 models. frequency of alcohol use.
16.92 12 .995 .976 .036 .703ns
Model 1b Path analytic model of units consumed in a single occasion.
168
Model 4f Path analytic model predicting binge recorded.
32.23 18 .984 .921 .065 .226ns
6.8 Chapter Summary
This chapter has presented the findings from the Phase I of mixed methods research, correlations between the variables in the study and the path analytic models were presented to address the hypotheses of the study. The path models presented in this chapter have predicted variance in intention 22% to 33% and the behaviour 22% to 55%. The findings from the Phase I of mixed methods study was successful in identifying significant contributors of alcohol consumption. In addition, the results assisted in identifying the components, which were significant in predicting alcohol consumption in the student population.
169