Chapter three.

Analytical results

3.1 Objectives of the analysis

The main goal of the current analysis is to firstly compare if there is a significant difference of

level of SAT components: morality, self-control, deterrence and peer drug-taking between drug

users and non-drug users, and between first time drug users and repeated drug users. To address

this question, a series of independent t-test were conducted. Secondly, in order to answer which

component of SAT has greatest predictability towards drug use behaviour, linear regression

between drug use frequency and four components was modelled. Thirdly, regarding the

question of whether the effect of self-control, deterrence and peer drug use are dependent on

morality, several interaction analysis were conducted in accompany of logistic regression to

investigate if participants’ identity (drug user or not) can be predicted by different level

components. This chapter presents the main finding of three types of analysis.

Firstly, a zero-order correlation was calculated to understand the relationships among the

variables. Figure 1 and Figure 2 show the correlation matrix between the four variables for all

respondents (including drug users, non-drug users, first-time detainees and repeated detainees).

The null hypothesis is that all variables are linearly independent. Two pairs of correlations can

be easily found with p-value smaller than 0.05.

Table 1 shows that there is a significant positive correlation between morality and Self-control

(r=0.42, p<0.05); Second, there is a significant positive correlation between morality and

Deterrence (r=0.32, p<0.05); Third, it reveals that there is a significant negative correlation

(r=0.4, p<0.05) between morality and peer. Lastly, there is a significant positive correlation

between Self-control and Deterrence (r=0.14, p<0.05). There was a significant negative

correlation between deterrence and peers (r = -0.12, p< 0.05). However, there are no other

obvious correlations (see Table 1 below).

Table 1: Correlation matrix with drug users and non-drug uers

Table 2 shows that, first, there is a significant positive correlation between morality and self-

control (r=0.25, p<0.05); Second, there is a significant positive correlation between morality

and deterrence (r=0.38, p<0.05); Third, there is a significant positive correlation between self-

control and deterrence (r=-0.31, p<0.05). Other pairs of the correlation are not statistically

significant at level of 0.05 (see Table 2 below).

Table 2: Correlation matrix with first-time detainees and repeated detainees

3.2 Independent Sample t-test

In order to compare the difference between each group, the independent sample t-tests are

introduced with respect to the impact levels of morality, self-control, deterrence and peer drug

usage on drug use behaviour between two groups of participants. Before conducting the main

analysis, a preliminary analysis is conducted to ensure that all assumptions are met. The sample

size for preliminary analysis is 166. For the main analysis, an independent sample t-test was

conducted with the null hypothesis that variables have the same mean in order to compare levels

of morality in drug users and non-drug users.

The results showed that the difference between the drug addict group (M = 31.89, SD = 7.96)

and the non-drug group (M = 51.96, SD = 17.76) are statistically significant at level of 0.05; t

(305.65) = 14.65, p < 0.001 (see Table 3 below).

At the same time, in the scale of self-control, the difference between the drug user group (M=

51.5, SD=18) and the non-drug user group (M=47.69,SD.=18.64): t (375) =-1.1, p < 0.05) are

statistically significant at level of 0.05. Also, in the scale of deterrence and peer drug usage, the

difference between the two groups is statistically significant at level of 0.05. First for

‘deterrence’, the output shows that (M= 29.6, SD =10.22) for the drug user group, and

(M=33.12, SD=13.52), t (374.41) =2.88, p < 0.05 for the non-drug user group. Second for peer

drug usage, the results were (M= 4.80, SD=3.2) for the drug user group and (M=1, SD=0),

t(165)=-15.3, p < 0.001 for the non-drug user group which is statistically significant at level of

0.05.
Table 3. Two-sample t-test (non-drug users vs. drug users)

Means comparison in morality, self-control, deterrence and peer drug use. Significant differences are

found in these four variables

Besides the comparison made between drug users and non-drug users, the same analysis was

conducted to compare the level of impact of each four SAT factors between first-time drug-

users and repeated drug users. In the scale of morality, a significant difference was found

between first-time-drug-users (M=34.45, SD=8.57) and repeated drug users (M=29.13,

SD=6.21), where t(154.98)=4.611，p < 0.001,which is statistically significant at level of 0.05.

In the meantime, in the scale of peer drug use, the difference between the first drug users (M =

3.43, S.D. = 2.97) and the repeat drug users (M = 6.28, SD = 2.76), t (164) =-6.37, p <0.001 are

statistically significant at level of 0.05. Finally, the difference between self-control deterrence

and peer drug use are not statistically significant at level of 0.05 (see Table 4 below).
Table 4. Two-sample t-test (first-time detainees vs. repeated detainees)

T-test table: means comparison in morality, self-control, deterrence and peer drug use between users in

first-time detainees vs. repeated detainees.

3.3 Linear Regression Analysis

Simple linear regressions are carried out to investigate how SAT’s four components: morality,

self-control, deterrence and peer drug usage behaviours influence participants’ frequency of

drug use. It has found that there is positive linear relationship between morality and drug use

frequency, which is confirmed with a Pearson’s correlation coefficient of 0.92, p<0.05 . Commented [YMY1]: Add the corresponded table behind

Four independent variables are taken as explanatory variables for the simple linear regression

respectively. This showed a significant relationship between morality and drug use (p < 0.001).

The R² value is 0.85, so 85% of the variation in drug use frequency can be explained by the

model containing morality (see Table 5.2 below). In terms of other variables, self-control,

deterrence and peer drug usage are not found to be significantly predictive towards drug use

frequency at level of 0.05. This approves that morality is the strongest predictor towards drug

use frequency (see Table 5 below).

Table 5. Linear regression of drug use frequency and SAT components

Table 5.1 Model Summaryb

Table 5.2 The change statistics of the linear regression between drug use frequency and four

components

A simple linear regression model is used to investigate the extent to which the components of the SAT

affect the drug users' drug use frequency (within 12 months).

A multiple linear regression analysis is performed to investigate the interaction effect between

the four components of the SAT (morality, self-control, deterrence, and concomitant drug use

behaviour) and the frequency of drug use among participants. It is found that there is a positive

linear relationship between moral and self-control interactions, moral and deterrent interactions,

and drug abuse frequencies, with a Pearson correlation coefficient of 0.92, p<0.05. The

estimation of coefficient for Morality is statistically significant at level of 0.05, which is 0.96

indicating that the with one unit’s increase in Morality, it would cause 0.96 unit’s increase in

frequency (see Table 6 below).

Also, multiple linear regressions are performed by using three independent variables. This

suggests a significant relationship between moral and self-control interactions and drug use (p

< 0.001). There was a significant relationship between morality and deterrence interaction and

drug use (p < 0.001). The R2 value is 0.54, so 54% of the frequency of drug use can be explained

by an ethical model (see Table 7.2). For the other variables, no significant predictive effect

of ethics and surrounding peers on the frequency of drug use is found.

Table 6. Linear Regression with the dependent variable of drug use frequency

Table 7. Interactive linear regression with the dependent variable of frequency of drug use

Table 7.1 ANOVAa

Table 7.2 Model Summaryb

After considering the interaction between morality and self-control, morality and deterrence

and morality and peer drug use, it is found that the coefficient of moral* self-control and

morality* deterrence are significant at level of 0.05 with corresponding estimation of

coefficient equalling to 0.004 and 0.007 (see Table 8).

Table 8. Linear regression with interaction (dependent variable: frequency of drug use)

3.4. Logistic Regression Analysis

In order to determine how SAT’s composing factors influence drug use patterns, two issues

were investigated. The first one is that whether these factors increase the likelihood of becoming

a drug user, and the other one is whether SAT’s components increase the likelihood of

becoming a repeated detainee.

To answer the first question, a three-predictor logistic model was built to test whether the levels

of morality, self-control and deterrence had significant impact on participants’ choice of using

drugs, and whether the effect of self-control and deterrence were dependent on morality levels.

Since none of the non-drug users had taken drugs, their data was excluded in the logistic

regression. The dependent variable used in this model was the category of participants’ type

based on the drug use behaviour (drug user / non drug user), which is a dummy variable. The

explanation variables are morality, self-control and deterrence.

The statistical significance of individual regression coefficients (βs) is tested using the Wald

chi-square statistic (Table 10). The chi-square shows that both morality and self-control were

significant predictors of individuals` drug use behaviour referrals for drug-users group and non-

drug users group (p < .05). The test of the intercept (i.e., the constant in Table 10) suggests

whether an intercept should be included in the model. For the present data set, the test result (p

< .05) suggested that an alternative model with the intercept might be applied to the data.

Goodness-of-fit statistics assess the fit of a logistic model against actual outcomes (i.e., whether

a referral is made for drug-user`s group). One inferential test and two descriptive measures are

presented in Table 9. The inferential goodness-of-fit test is the Hosmer–Lemeshow (H–L) test

that yielded a χ2(8) of 47.31 and was significant (p < .05), suggesting that the model was fit to

the data well.

Table 9. Logistic Regression

Table 10. Logistic Model without interaction (Dependent variable: drug user=2; non-drug
user=1)

In addition, to test whether the effects of self-control, deterrence, and peers depend on morality

(ie, whether these factors interact with morality), morality* self-control, morality* deterrence

and morality* peer drug use are used as independent variables. The results show that the

interaction effect between influence of self-control and morality is not statistically significant

at level of 0.05. Thus, the results show that none of the interaction between these variables is

statistically significant at level of 0.05 (see Table 11 and 12).

The inferential goodness-of-fit test is the Hosmer–Lemeshow (H–L) test that yielded a χ2(6) of

21.64 and was significant (p < .05), suggesting that the model was fit to the data well (Table

11).

Table 11. Logistic Regression

Model Summary

Step -2 Log likelihood Cox & Snell R Square Nagelkerke R Square

1 84.40a .68 .92

a. Estimation terminated at iteration number 20 because maximum iterations has been reached.

Final solution cannot be found (see Table 12).

Table 12. Logical models with interactions (interaction) (dependent variables: drug users = 2;

non-drug users = 1)

To answer the second question, again, the three predictive logistic models are used to test

whether morality, self-control, and deterrence levels have any effect on the choices of becoming

a repeat detainee, and whether self-control and deterrence depend on morality. The dependent

variable in this model is repeated drug abuse behaviour, with explanatory variables are morality,

self-control, deterrence, and peers.

The statistical significance of individual regression coefficients (βs) was again tested using the

Wald chi-square statistic (Table 14). The chi-square shows that both morality and peer drug use

were significant predictors of individuals` drug use behaviour referrals for first time drug users

group group and repeated drug users group (p < .05). The test of the intercept (i.e., the constant

in Table 14) (p > .05) suggested that an alternative model without the intercept might be applied

to the data.

In terms of the goodness-of-fit statistics, one inferential test and two descriptive measures are

presented in Table 9. The inferential goodness-of-fit test is the Hosmer–Lemeshow (H–L) test
that yielded a χ2(6) of 16.16 and was significant (p < .05), suggesting that the model was fit to

the data well.

Table 13. Logistic Regression

Table 14. Logistic Model without interaction (Dependent variable: repeated drug use behaviour;

repeated=2; first-time=1)