We hypothesized that: a) motives that promote positive experiences would not be associated with symptoms of depression, symptoms of anxiety, or overall psychological distress; b) motives for avoidance of negative experiences would be associated with higher levels of symptoms of depression and symptoms of anxiety, or overall psychological distress; c) motives focused on medicinal use would be associated with lower levels of symptoms of depression and symptoms of anxiety, or overall psychological distress; and d) there would be no association between motives of boredom, relative low risk, and availability with depression or anxiety symptoms of depression and symptoms of anxiety, or overall psychological distress. As a first step, multiple linear regression analyses were used to investigate the associations between motives of marijuana use and symptoms of depression and symptoms of anxiety as well as overall psychological distress in our sample. Variables were entered in two blocks using the “enter” function for regressions in SPSS. The first block consisted of the 17 motives of use and the second block entered contained the control variables: age, race/ethnicity, user group, and gender. Given the number of variables entered in the model and the number of comparisons to be made, Bonferroni corrections were used to counteract potential Type I errors. Thus, the Bonferroni corrected alpha value of 0.003 was used to assess significance. Post hoc power analyses, or the probability of finding a statistical difference from zero,vertical grow room design were also performed. Second, mediation analyses using a non-parametric bootstrapping approach were conducted to assess whether past 90 days marijuana use or daily number of marijuana hits influenced the association between motives of marijuana use and mental health in our sample.
The mediation analyses followed PROCESS Model 4 . A cross product test of the coefficients was favored over causal step mediation as it is a superior method to detect indirect effects and assess their significance . The cross product of the coefficients test provides a single test for the relation between the independent variable, the mediator, and the dependent variable by multiplying coefficients for a and b paths, therefore directly assessing the statistical significance of the indirect effect using bootstrapped confidence intervals. Testing the cross product of coefficients using a nonparametric bootstrapping method is advantageous as it does not require for the assumption of normality to be met, and is appropriate for smaller to moderate sample sizes . To assess for significant indirect effects, 95% bias corrected confidence intervals were calculated using 10,000 bootstraps. Indirect effects were considered significant if the 95% bias corrected confidence intervals for ab point estimates did not contain zero . To further correct for Type I errors, a supplemental analysis using 99% bias corrected confidence intervals were also calculated using 10,000 bootstraps. To better quantify and compare the effect size of each indirect effects, completely standardized effects were calculated . Completely standardized effects express the indirect effects as the change in the standard deviation for the dependent variable between two cases of the independent variable that differ by one standard deviation . Analyses were conducted using Version 3 of the PROCESS macro in SPSS Version 24, first without any control variables and subsequently controlling gender, age, user group, and race/ethnicity. Men, non-patient users, and Non-Hispanic Whites were used as reference categories for gender, user group, and race/ethnicity respectively. The purpose of this third aim was to determine whether associations between motives of use and our mental health outcomes of interest varied by gender. First, moderation analyses were performed to examine whether the associations between motives of marijuana use and symptoms of depression, symptoms of anxiety, and overall psychological distress differ by gender in young adults who use marijuana. Second, conditional process analyses were done to test for gender differences for the significant indirect associations between motives of marijuana use and mental health outcomes uncovered in aim 2.
Men was used as the reference category for all moderation and conditional process analyses. Analyses were performed using the PROCESS Version 3 macro in SPSS Version 24. PROCESS Model 1 was used to assess moderation . Per Hayes , a moderation is deemed significant if the coefficient for the interaction term between the independent variable and the moderator is significant. In this scenario, the coefficient will properly estimate the moderation of the independent variable’s effect by the moderator . An interaction term was deemed significant if p ≤ 0.05. Conditional process analyses, also called moderated mediation, were conducted to determine whether gender influences the indirect effects found to be significant in aim 2. In these moderated mediation models, the strength of the relationship between motives of marijuana use on symptoms of depression, symptoms of anxiety, or psychiatric distress is conditional on the value of the moderator; gender. Given that our interest was to test the effect of gender on the three paths of the mediated model X→M, M→Y, X→Y, Hayes’ PROCESS Model 59 was used for the conditional process analyses . By using this model, a test of moderation for each path is available in the form of the regression coefficients for the products along with their tests of significance. PROCESS also generates tests of significance and bootstrapped confidence intervals for the conditional direct and indirect effects. PROCESS also automatically conducts a test of the difference between the indirect effects in the two groups called the index of moderated mediation, with a bootstrapped confidence interval. The index of moderated mediation and its bootstrap confidence interval therefore act as an inferential test for the conditional process analysis of the indirect effect . In summary, by conducting conditional process analyses using PROCESS Model 59, we were able to determine which path, if any, was significantly moderated, and whether the indirect effect was moderated.
Bootstrapped confidence intervals for the conditional indirect effects were calculated using 10,000 bootstraps. Using bootstrapped confidence intervals can help avoid power problems introduced by asymmetric and other non-normal distributions of an indirect effect . Descriptive statistics for the sample are presented in Table 3.4. Two cases were eliminated from the original dataset as their gender identity was defined as “other”. Participants were on average 21 years old and mostly men . Forty-five percent of respondents identified as Hispanic/Latino, 26% as Non-Hispanic White, and 19% as NonHispanic African American/Black, 4% as Asian/Pacific Islander, and 6% as multi-racial. This racial/ethnic distribution is somewhat comparable to that of Los Angeles County . Past year annual income was relatively low with 83% of the sample falling in the $1-$25,000 bracket. Most participants reported part-time employment. With regards to education, about half of the sample reported having completed some college and/or being currently enrolled in either a four year or community college. Marijuana was the most frequently used drug in the past 90 days. On average, participants reported using marijuana 69 out of the past 90 days. This means that, on average, participants used marijuana between on 5 to 6 days per week,grow vertical thus classifying their use as heavy . Use of heroin was only reported by one participant over the past 90 day period. The average daily number of marijuana hits was 23.5. There was no difference between men and women with regards to either past 90 days use or daily number of hits. Overwhelmingly, participants reported smoking buds/flowers as the primary form and way of marijuana use. On about 26 of the past 90 days, marijuana was used with other drugs, primarily alcohol about 43% of the time. Fifty-seven percent of the sample had a valid medical marijuana recommendations and thus identified as medical marijuana patients or medical marijuana users. With regards to motives of use, the motive of enjoyment was the motive with the highest mean score indicating that “most of the time” participants in the sample used marijuana for enjoyment purposes . This is followed by motives of sleep and relative low risk. When examining the mode of motives , “always” is the most frequent answer for motives of sleep, relative low risk, pain, and enjoyment. Motives of altered perceptions, availability and celebration follow with “most of the time”. There was a significant difference in mean scores of reported motives of use between men and women for motives of attention, celebration, enjoyment, natural remedy, nausea, pain, sleep and social anxiety . For all these motives, women scored higher than men. Brief Symptoms Inventory-18 scores averaged between 3 and 4 out of a possible 24 for both symptomatology of depression and symptomatology of anxiety, indicating that participants in our sample endorsed some symptoms of depression or anxiety. For the Global Severity Index, which is used to operationalize psychological distress, the average score for the sample was 9.89 out of a possible 72. Only for the symptomatology of anxiety and psychological distress scales was there a significant difference of scores by gender . Table 4.41 presents the regression estimates of symptoms of depression on motives of marijuana use without and with control variables. Motives of use account for 22% of the variance in symptoms of depression. At p ≤ 0.05, motives of celebration, coping and pain were significantly associated with symptoms of depression in the analyses without control variables.
After controlling for age, gender, race/ethnicity, and user group, only coping remained significantly associated with symptoms of depression. At a Bonferroni corrected alpha of ≤ 0.003., only coping was positively, significantly associated with symptoms of depression in models without and with control variables. None of the control variables included in the model were significantly associated with symptoms of depression. The association between the coping motive of marijuana use with symptoms of depression is positive indicating that the more often marijuana use is motivated by coping, the higher the score for symptoms of depression. The magnitude of changes in symptoms of depression for a one unit increase in motives of use is of almost 2 points. Post hoc power analyses indicate that the statistical power is greater than 0.9. Results from the mediation analysis with past 90 days marijuana use as a mediator are presented in Tables 4.42a-d. From a simple mediation analysis without control variables , marijuana use motives of availability, conformity, pain, and social anxiety indirectly influenced symptoms of depression through their effect on past 90 days marijuana use. For motives of availability and conformity, the indirect association through past 90 days use is positive , whereas it is negative for motives of pain and social anxiety . For each of these indirect effects, a 95% bootstrap confidence interval based on 10,000 bootstraps did not contain zero . For motives of conformity, coping, and social anxiety, there is also evidence of a direct effect with symptoms of depression independent of their effect on past 90 days marijuana use . The effect is positive for motives of coping and social anxiety with symptoms of depression whereas the direct effect between conformity and symptoms of depression is negative. After controlling for age, gender, race/ethnicity, and user group , the indirect effect of motives of availability on symptoms of depression and social anxiety on symptoms of depression through past 90 days use were no longer significant. Significant indirect effects remained for the motives of conformity and pain with symptoms of depression. For each of these indirect effects, a 95% bootstrap confidence interval based on 10,000 bootstraps did not contain zero . The completely standardized effect for the motive of pain was of -0.26 and of 0.22 for the motive of conformity. Evidence of a direct effect remained for the motive of social anxiety with symptoms of depression but not for the availability motive. The a path from motive of conformity to past 90 days marijuana use was negative, indicating that the more use is driven by conformity , the less days one is likely to use. However, for motive of pain the association was positive, indicating that the more use is driven by this motive, the more days of use is reported. Motives of use accounted for 19% of the variance of past 90 days marijuana use. Past 90 days of marijuana use was significantly, yet negatively, associated with symptoms of depression. However, although significant, the magnitude of the b coefficient here was almost 0. For each of these indirect effects, a 95% bootstrap confidence interval based on 10,000 bootstraps did not contain zero . Table 4.44 presents the multiple linear regression estimates without and with control variables.