This is an introduction for the second course of Data Analysis and Interpretation Specialization, offered by Wesleyan University through Coursera. For the grading purposes of the course, the assignments were initially uploaded on Tumblr.
It is an attempt to develop and test hypotheses with a variety of statistical tools, such as ANOVA, Chi-Square Test and Pearson Correlation analysis, working on existing data (U.S. National Epidemiological Survey on Alcohol and Related Conditions (NESARC).
For the code and the output i used Spyder (IDE). Requires python 2.7+.
- Hypothesis Testing and ANOVA
- Chi Square Test of Independence
- Pearson Correlation
- Exploring Statistical Interactions
The data was provided by the National Epidemiological Survey on Alcohol and Related Conditions (NESARC), which was conducted in a random sample of 43,093 U.S. adults and designed to determine the magnitude of alcohol use and psychiatric disorders. Sample size is important because the larger the sample size, the more accurate the findings. NESARC’s unusually large sample size also made it possible to achieve stable estimates of even rare conditions. NESARC participants came from all walks of life and a variety of ages, and the level of analysis studied was individual. They represented all regions of the United States and included residents of the District of Columbia, Alaska, and Hawaii. In addition to sampling individuals living in traditional households, NESARC investigators questioned military personnel living off base and people living in a variety of group accommodations such as boarding or rooming houses and college quarters. More specifically, the sample consists of 24,575 (57.1%) males and 18,518 (42.9%) females, among of whom 9,535 (22.13%) were aged between 18 and 30 years old. The data analytic subset, examined in this study, includes individuals aged between 18 and 30 years old who reported using cannabis at least once in their life (N=2,412).
In 2001—2002, the National Institute on Alcohol Abuse and Alcoholism (NIAAA) conducted the first wave of the National Epidemiological Survey on Alcohol and Related Conditions (NESARC), the largest and most ambitious survey of this type conducted to date. Information was collected in face-to-face computer-assisted interviews, which took place in the participants’ homes. It contained an extensive battery of questions about present and past alcohol consumption, AUDs, and the use of alcohol treatment services. NESARC also included similar questions related to tobacco and illicit drug use (including nicotine dependence and drug use disorders), as well as questions designed to determine a wide variety of psychiatric disorders such as major depression, anxiety disorders, and personality disorders. The original purpose of this survey was to evaluate the magnitude and have a better understanding of the link between alcohol use and other drug use and/or psychiatric disorders, which can help treatment providers design more targeted screening and more effective treatments for their patients. The response rate was 81%, which is significantly high compared to the standards of recent large-scale national surveys. A high response rate is very important, since it is key to legitimizing the results of the survey.
Major depression and general anxiety diagnoses (categorical response variables), diagnosed in the last 12 months, were assessed using DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition), a manual published by the American Psychiatric Association (APA) that includes all currently recognized mental health disorders. Diagnoses’ data measures were drawn from NESARC codebook, in DSM-IV section. In addition, the ‘medicine use’ section of the codebook includes detailed information about the frequency (categorical explanatory variable) and the quantity (quantitative explanatory variable) of cannabis use. Current cannabis use was evaluated through both frequency (”How often used cannabis when using the most?”), which is a categorical variable binned into 10 categories, and quantity (”Number of joints usually smoked in a day when using cannabis the most.”), which is a quantitative variable that ranged from 1 joint per day to 98 joints per day.
The response format of the explanatory variable (frequency of cannabis use) was a ten-point scale, since variable was binned into 10 categories (1.“Every day”, 2.”Nearly every day”, 3.”3-4 times a week”, 4.“1-2 times a week”, 5.“2-3 times a month”, 6.“Once a month”, 7.“7-11 times a year”, 8.“3-6 times a year”, 9.“2 times a year”, 10.“Once a year”) and participants could choose an answer among these response options. As far as major depression and general anxiety diagnoses (response variables) are concerned, they were coded dichotomously (two-point scale), which indicates options that were absolutely opposite to each other (1.”Yes”, 2.”No”).
In order to evaluate the magnitude of the current cannabis use, smoking quantity was taken into consideration. Thus, from the quantity of joints smoked per day when using the most (quantitative variable), a new secondary variable was created (categorical variable), that estimates on average the quantity of joints smoked per month, by multiplying the number of joints smoked per day with the number of days an individual smoked per month.
An ANOVA (C->Q) was used, for the examination of the relationship between the psychiatric disorders (categorical explanatory variables) and cannabis use quantity (quantitative response variable).
For the correlation between frequency of cannabis use (explanatory variable) and such psychiatric disorders (response variables), a Chi-square Test of Independence (C->C) was used, so that the chi-square and p values were measured. Furthermore, in order to determine which frequency groups are different from the others, post hoc pair comparisons were performed, using Bonferroni Adjustment approach, since the explanatory variable had more than 2 levels. In addition, the association was visualized graphically, using a bivariate bar chart.
Subsequently, the hypothesis was refined and Pearson Correlation analysis (Q->Q) was used, to examine the association between the age when the individuals began using cannabis the most (quantitative explanatory variable) and the age when they experienced the first episode of major depression and general anxiety (quantitative response, variable). The results were also visualized with scatterplots.
Last but not least, a categorical two-point scale (1.”Yes”, 2.”No”) lurking variable (”Any family members or close friend died in the last 12 months.”) was taken into account, in order to examine if this factor moderates the association between cannabis use and both major depression and general anxiety diagnoses in the last 12 months.