Data Management and Visualization @nmansoorn - Tumblr Blog

Moderation

Moderation in The Context of Correlation

Now that I know how to take into account a third variable, a particular questions that I needed answer for was income level is somehow associated as a third variable on my previous analysis in discovering whether there is a relationship between drinking and gambling. For this I had to categorize the income level into 5 categories as follows:

7=less than or equal to 9,999;

15=10,000-19,999;

28=20,000-34,999;

45=35,000-59,999;

75=60,000-89,999;

105=90,000-119,999;

160=120,000+

For the purpose of this test I am going to run a CORRELATION COEFICIENT test on the 7 income categories that I have created.

The results show that for 2 the lowest income group, as I expected, does not show a significant value both in its correlation coefficient and its P-value. However, this relation gets really strong as the income level rises. So with this information we can safely conclude that income level has a significant impact as a moderator to determine the direction of the study, and define whether the person who had a lot drinks is also going to gamble or not. If they are on the upper levels of income group, they will, if they are not they won't.

-----------------------

Code Screenshot:

-----------------------

Code Text:

-----------------------

LIBNAME mydata "/courses/d1406ae5ba27fe300" access=readonly; DATA new; set mydata.nesarc_pds; LABEL S1Q11A="Total Family Income in Last 12 Months - Category" S2AQ7D="Number of Drinks of Liquor Consumed on Days When Drunk" S12Q1="Ever Gambled 5+ Times in Any One Year" DGSTATUS="Drug Use Status" S1Q6A="Highest Grade of Year of School Completed" SPOUSE="Respondent Has a Spouse" S3BQ1A9A="Ever Used Heroin" S3BD9Q2A="Age First Used Heroin" S3BD9Q2B="Used Heroin in Last 12 Months" NUMPERS="Number of Persons in Household" S3BD5Q2E="How Often Used Cannabis When Using the Most" S3BD9Q2C="How Often Used Heroin in Last 12 Months" S3BD9Q2E="How Often Used Heroin When Using the Most" AGEGROUP="Age First Used Heroin"; /* Just giving AGEGROUP a lable too */ IF S12Q1 LE 2; IF S3BQ1A9A=9 THEN S3BQ1A9A=.; IF S3BQ1A9A NE 9 AND S3BD9Q2B=. THEN S3BD9Q2B=11; IF S3BQ1A9A NE 9 AND S3BD9Q2A=. THEN S3BD9Q2A=99; /* Assiging 99 to avoide misrepresentation */ IF S3BD9Q2A NE 99 AND S3BD9Q2A LE 17 THEN AGEGROUP=1; /* Ages 17 and below as group 1*/ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 18 AND S3BD9Q2A LE 30 THEN AGEGROUP=2; /* Ages 18 - 29 as group 2 */ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 31 THEN AGEGROUP=3; /* Ages 30 and above as group 3 */ IF S3BQ1A9A NE 9 AND S3BD9Q2E=. THEN S3BD9Q2E=11; IF DGSTATUS LE 2; /* Retriving data only for current or ex- drug users */ IF AGE LE 45; /* Intrested in the age group of 45 and below */ IF S2AQ7D=99 THEN S2AQ7D=.; IF S3BD5Q2E=99 THEN S3BD5Q2E=.;

/*USINC: 7=less than or equal to 9,999; 15=10,000-19,999; 28=20,000-34,999; 45=35,000-59,999; 75=60,000-89,999; 105=90,000-119,999; 160=120,000+ */

IF S1Q11A LE 9999 THEN USINC =7; IF S1Q11A GE 10000 AND S1Q11A LE 19999 THEN USINC=15; IF S1Q11A GE 20000 AND S1Q11A LE 34999 THEN USINC=28; IF S1Q11A GE 35000 AND S1Q11A LE 59999 THEN USINC=45; IF S1Q11A GE 60000 AND S1Q11A LE 89999 THEN USINC=75; IF S1Q11A GE 90000 AND S1Q11A LE 119999 THEN USINC=105; IF S1Q11A GE 120000 THEN USINC=160; /* PROC SORT; by USINC;

PROC FREQ; TABLES S12Q1*S2AQ7D/CHISQ; BY USINC; */

PROC SORT; BY USINC; PROC CORR; VAR S12Q1 S2AQ7D; BY USINC;

RUN;

#moderation #coursera

Correlation Coefficient

Two the things that I though might have effect on how often people use cannabis were; first, how drunk they are, and second; the how many people are there in their household. This is because I thought, the more people drink, the more they are willing to use cannabis, and on the other hand, the more people are there in their households, the more likely they are not to smoke due to respect and inability.

To examine this, I checked the correlation coefficient for both of these questions in relation to how often they used cannabis.

Unlike my expectation, the results reveal another story. It looks like even though the number of drinks have an impact on the person’s decision to use cannabis, however, this relationship is negative. Meaning that, they more the person drink, the less likely he is to use cannabis. The relationship is not quite strong though. With a value of (-0.172) and a P value of 0.001, this dulls the overall judgment on the subject.

The same could be implied to the second questions, because the R value is really close to zero (0.016). Even though this stands on the positive side, the relationship is pretty weak, with a P value of more than 0.05, and this again fails to prove a strong relationship between the values.

----------------------------------

Code Text:

----------------------------------

LIBNAME mydata "/courses/d1406ae5ba27fe300" access=readonly; DATA new; set mydata.nesarc_pds; LABEL DGSTATUS="Drug Use Status" S12Q1="Ever Gambled 5+ Times in Any One Year" S1Q6A="Highest Grade of Year of School Completed" SPOUSE="Respondent Has a Spouse" S3BQ1A9A="Ever Used Heroin" S3BD9Q2A="Age First Used Heroin" S3BD9Q2B="Used Heroin in Last 12 Months" S2AQ7D="Number of Drinks of Liquor Consumed on Days When Drunk" NUMPERS="Number of Persons in Household" S3BD5Q2E="How Often Used Cannabis When Using the Most" S3BD9Q2C="How Often Used Heroin in Last 12 Months" S3BD9Q2E="How Often Used Heroin When Using the Most" AGEGROUP="Age First Used Heroin"; /* Just giving AGEGROUP a lable too */ IF S3BQ1A9A=9 THEN S3BQ1A9A=.; IF S12Q1=9 THEN S12Q1=.; IF S3BQ1A9A NE 9 AND S3BD9Q2B=. THEN S3BD9Q2B=11; IF S3BQ1A9A NE 9 AND S3BD9Q2A=. THEN S3BD9Q2A=99; /* Assiging 99 to avoide misrepresentation */ IF S3BD9Q2A NE 99 AND S3BD9Q2A LE 17 THEN AGEGROUP=1; /* Ages 17 and below as group 1*/ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 18 AND S3BD9Q2A LE 30 THEN AGEGROUP=2; /* Ages 18 - 29 as group 2 */ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 31 THEN AGEGROUP=3; /* Ages 30 and above as group 3 */ IF S3BQ1A9A NE 9 AND S3BD9Q2E=. THEN S3BD9Q2E=11; IF DGSTATUS LE 2; /* Retriving data only for current or ex- drug users */ IF AGE LE 45; /* Intrested in the age group of 45 and below */ IF S2AQ7D=99 THEN S2AQ7D=.; IF S3BD5Q2E=99 THEN S3BD5Q2E=.;

PROC SORT; by IDNUM;

PROC CORR; VAR S2AQ7D NUMPERS S3BD5Q2E;

RUN;

------------------------------

#coursera #SAS

Chi Square Test

Here I am examining whether drug usage status is associated with gambling habit.

H0: Gambling is not associated with status of drug use

H1: Gambling is associated with status of drug use

Since the P value is 0.0024, we can conclude that status of drug use is associated with gambling. That is, if the person has drugs ever, or in the past 12 months, this raises the chances that they have also gambled. With this in mind, we can reject the null hypothesis saying there is no association between drug status and gambling, and accept the alternate hypothesis.

The Tables:

The Code Screenshot:

The Code Text:

-----------------------------------------------------------------------------------

LIBNAME mydata "/courses/d1406ae5ba27fe300" access=readonly; DATA new; set mydata.nesarc_pds; LABEL DGSTATUS="Drug Use Status" S12Q1="Ever Gambled 5+ Times in Any One Year" S1Q6A="Highest Grade of Year of School Completed" SPOUSE="Respondent Has a Spouse" S3BQ1A9A="Ever Used Heroin" S3BD9Q2A="Age First Used Heroin" S3BD9Q2B="Used Heroin in Last 12 Months" S3BD9Q2C="How Often Used Heroin in Last 12 Months" S3BD9Q2E="How Often Used Heroin When Using the Most" AGEGROUP="Age First Used Heroin"; /* Just giving AGEGROUP a lable too */ IF S3BQ1A9A=9 THEN S3BQ1A9A=.; IF S12Q1=9 THEN S12Q1=.; IF S3BQ1A9A NE 9 AND S3BD9Q2B=. THEN S3BD9Q2B=11; IF S3BQ1A9A NE 9 AND S3BD9Q2A=. THEN S3BD9Q2A=99; /* Assiging 99 to avoide misrepresentation */ IF S3BD9Q2A NE 99 AND S3BD9Q2A LE 17 THEN AGEGROUP=1; /* Ages 17 and below as group 1*/ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 18 AND S3BD9Q2A LE 30 THEN AGEGROUP=2; /* Ages 18 - 29 as group 2 */ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 31 THEN AGEGROUP=3; /* Ages 30 and above as group 3 */ IF S3BQ1A9A NE 9 AND S3BD9Q2E=. THEN S3BD9Q2E=11; IF DGSTATUS LE 2; /* Retriving data only for current or ex- drug users */ IF AGE LE 45; /* Intrested in the age group of 45 and below */ PROC SORT; by IDNUM; /*PROC ANOVA; CLASS DGSTATUS; MODEL SPOUSE=DGSTATUS; MEANS DGSTATUS/DUNCAN;*/

PROC FREQ; TABLES S12Q1*DGSTATUS/CHISQ;

RUN;

------------------------------------------------------------------------------------------

#coursera #chisquare #SAS

Hypothesis and ANOVA

The questions I want an answer for is whether having spouse is somehow related to drug addiction. Although in this question both variables are categorical, I am still using it to examine how the tool works (as allowed by the assignment).

Following are the hypothesis:

H0: No relation between drug usage status and having a spouse.

HA: Strong relation between drug usage status and having a spouse.

As it can be seen from the results, the P-Value is 0.0001, which means there is a strong relationship between having a spouse and the status of drug use.

Below is the text for the code:

-------------------------------------------

LIBNAME mydata "/courses/d1406ae5ba27fe300" access=readonly; DATA new; set mydata.nesarc_pds; LABEL DGSTATUS="Drug Use Status" S1Q6A="Highest Grade of Year of School Completed" SPOUSE="Respondent Has a Spouse" S3BQ1A9A="Ever Used Heroin" S3BD9Q2A="Age First Used Heroin" S3BD9Q2B="Used Heroin in Last 12 Months" S3BD9Q2C="How Often Used Heroin in Last 12 Months" S3BD9Q2E="How Often Used Heroin When Using the Most" AGEGROUP="Age First Used Heroin"; /* Just giving AGEGROUP a lable too */ IF S3BQ1A9A=9 THEN S3BQ1A9A=.; IF S3BQ1A9A NE 9 AND S3BD9Q2B=. THEN S3BD9Q2B=11; IF S3BQ1A9A NE 9 AND S3BD9Q2A=. THEN S3BD9Q2A=99; /* Assiging 99 to avoide misrepresentation */ IF S3BD9Q2A NE 99 AND S3BD9Q2A LE 17 THEN AGEGROUP=1; /* Ages 17 and below as group 1*/ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 18 AND S3BD9Q2A LE 30 THEN AGEGROUP=2; /* Ages 18 - 29 as group 2 */ IF S3BD9Q2A NE 99 AND S3BD9Q2A GE 31 THEN AGEGROUP=3; /* Ages 30 and above as group 3 */ IF S3BQ1A9A NE 9 AND S3BD9Q2E=. THEN S3BD9Q2E=11; IF DGSTATUS LE 2; /* Retriving data only for current or ex- drug users */ IF AGE LE 45; /* Intrested in the age group of 45 and below */ PROC SORT; by IDNUM; PROC ANOVA; CLASS DGSTATUS; MODEL SPOUSE=DGSTATUS; MEANS DGSTATUS; "*/ PROC FREQ; TABLES /*DGSTATUS S3BQ1A9A S3BD9Q2A AGEGROUP S3BD9Q2B S3BD9Q2E SPOUSE S3BD9Q2C*/; PROC GCHART; VBAR SPOUSE/Discrete type=PCT Width=30; PROC GCHART; VBAR S3BD9Q2C/type=PCT; PROC GCHART; VBAR SPOUSE/Discrete type=mean SUMVAR=S3BD9Q2C; PROC UNIVARIATE; VAR S3BD9Q2C; */" RUN;

--------------------------------------------------

#coursera #courser hypothesis and anova #hta

Finding a Relation Between Heroin Addiction and Relationship Status

The question that I am interested in at this point of time is whether heroin addiction is related to the respondent relationship status. I want to know whether having a spouse is directly or indirectly related to the respondent’s addiction.

Around 47 percent of all the respondent do not have spouses and 53 percent of them do. On the other hand, I pulled the graph for how often did these respondents used heroin in the last 12 months. Since it was already in small number of categories, I didn’t have to categories.

In this causal model, the explanatory variable is having a spouse and the dependent variable is frequency of heroin use. I was amazed by the results. I though heroin would mostly be used by single people, however the data shows something totally different. It actually has a direct relation with having a spouse.

#Coursera DMAV #dmav #coursera

Managing the Data

Besides some of the missing data available in the survey which needs to be set aside, there are also a lot of observations that are blank, but valid, and needs to be coded in and the values be recorded. For instance, for the question “Age first used drugs” there are more than 4,000 blank answers because it’s not applicable for those respondents who said they have never used drugs. Thus, there are a lot of “skip patterns” that need to be coded in appropriately and as valid data by replacing the values to “11”.

It’s going to work like a filter. So if someone has been a “lifetime nondrug user” that also means that he is she has never used heroin and thus a skip pattern is generated here. Following on that, if someone responds “No” to the question “Ever Used Heroin”, that means all the questions following this question about heroin use is going to be inappropriate, so another skip pattern is generated. The only missing data that needs to be excluded from the evaluation is the unknown responses on the question “Ever Used Heroin”.

I then categorized the age into three categories. Important thing to notice is there you can see age of “99”, which represents the people who have never used drugs. I replaced the “11” cod (used on other tables) with “99” to make sure it does not contradict the actual value of age “11”, and avoid misrepresentation. Accordingly, I am grouping the ages i.e. below 17 as group 1, 18-27 as group 2, and 28 above as group 3.

One of my findings from the survey is that all the respondents are aged 18 and above, so some the questions like “Age First Used Heroin” could not be answered with “AGE” variable – but rather the specific variable in question i.e. S3BD9Q2A. And since I am more interested in the categories of age rather than the knowing frequency for every single age, I am excluding the detailed age frequency from the results and just showing the answers in age groups. This way it makes more sense and looks more neat, organized and understandable.

#SAS #coursera #dmav

Week II Assignment

There are a lot of variable that I need to look at. First of all, I need to know at what age did these people start using any kind of drug, and then I’ll try to find if there is a relation between using drugs first and then doing heroin.

The data reveals some shocking results. The youngest drug users is shown to be 5 years old, even though the frequency is very low it is still shocking to know people use drug at that young age. Almost 22% of people start using drugs at the ages of 13-17, and another 30% are those aged between 18-21. Ages 22-30 constitute another 30% of drug users in the United States. This concludes that 82 percent of drug users are aged between 13-30 years. Based on sample of 43,006 people, however more than 42,000 of respondents skipped the question. On the other hand, common age for heroin starters is revealed to be 18-25, constituting more than 70 percent of heroin users.

When asked whether the used heroin in the last 12 months, only 2 percent of the respondents said yes, while around 94 percent said they were using it before the 12 months period – out of 150 total respondents.

When asked about how often were they using heroin the last 12 months, 36% of the respondents said they were using it everyday to almost everyday, and another 36% percent laid on the side saying they used only 2 times a year.

The next questions was, how often were they using heroin when they were using it the most, and similarly to the previous question, a large majority of almost 43% said they were using it everyday, while on the other side, almost 44% of the people said they were using it once to twice a year.

Lastly, I wanted to know at what age they started using the heroin the most. Running the data, I discovered that most of the people use heroin at its max when they are between the ages of 17-20, constituting more than 40 percent of the entire users, and 19 being the mode.

In concluding, I discovered that drug usage, regardless of what it is, including heroin, is at its peek at the ages of 17-20. I will look into this further on future assignments to discover why it’s used enormously at this age, and what are the possible ways to prevent it.

#SAS #Coursera DMAV

NESCAR Research Project

Heroin Addiction

After brainstorming numerous ideas and looking through the problem sets, I was still confused on what to problem to work on, that will keep me interested and I can actually learn from my findings. This took me several days, until I thought why not think about the problems my country is currently facing. With that in mind, I mentally positioned myself back on the streets of my home city, Kabul, Afghanistan, and looked around to see what problem do I immediately notice. While it wasn’t to come up with a huge list, one of the problems that is too common and that no one is actually working on it is excessive drug use, specifically, heroin. There are hundreds of men, located under the bridges and destroyed constructions injecting themselves and taking heroin by some other methods, which I have limited knowledge about at this point of time.

The question that comes to my mind, looking at all these people are, weather addiction to heroin is related to their income, and whether immigration has to do something with the addiction. This is because, most of them are poor and homeless, or maybe they became homeless because they spent everything on their drug. Also because most of these people had once migrated to some neighboring countries and brought this habit with them from there.

1. I chose NESCAR data set.

2. Topic of interest:

a. Heroin addiction

3. Variables:

a. Relation of heroin addiction with income

b. Relation of heroin addiction with migration

4. Code book pages:

a. NESCAR (107 – 290)

5. Literature Review

a. Although there wasn’t much written about the topic under discussion about Afghanistan, the following might be helpful to find a relation between the occurring issues.

i. http://www.sciencedirect.com/science/article/pii/S0376871602003423 ii. http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9566.2007.01018.x/full

b. Following research is really helpful in finding out the origins of the problem and the risks associated with it.

i. http://harmreductionjournal.biomedcentral.com/articles/10.1186/1477-7517-2-13

ii. http://www.organized-crime.de/revmcc01.htm

6. Hypothesis:

a. There is a strong relation between excessive drug use and local drug production.

b. There is a direct relation between income level, whether the person has migrated, and heroin addiction.

c. Peer pressure has had a huge impact on Afghans who has worked for the neighboring countries.

#coursera #dmav

Trending Blogs

Recently Viewed Blogs

Data Management and Visualization