Creating Graphs - Week 4

The two variables considered for the research question are graphed below using univariate (count) and bivariate bar charts. They are:
  • Response variable: Active Drinkers vs. Sober Respondents in a subset of the NESARC dataset that is limited to respondents who have sought at least one type of treatment for alcohol dependence.
  • independent variable: Count of distinct types of treatment for alcohol dependence sought by an individual respondent.
This univariate bar chart compares the two possible values of the response variable: Active vs. Sober Drinkers. Of the approximately 400 respondents who sought at least one type of treatment for alcohol dependence, about 100, or 25% remained sober for the 12 months prior to being surveyed.

 
 



This univariate bar chart shows the distribution of the independent variable: the number of distinct treatment types sought per respondent. This variable was computed by totaling the instances across the various types of treatments included in the NESARC survey for each respondent. This numerical total was then converted to a categorical variable.

The distribution is skewed right (long tail on the high side) so the values were then grouped in ways that approximate a normal distribution as shown in the chart below. The number of treatment types peaks at 2, then gradually trails off until 8. Values greater than 8 represent much smaller counts than the lower values. To account for this the last treatment count category includes 5 treatment count values (9, 11, 12, 13 & 14).

 
The following is a bivariate bar chart compares the independent variable, number of distinct treatment types per respondent, to the response variable, sober vs. active drinkers.
The chart suggests that the more distinct types of treatment a respondent has sought, the higher the likelihood that they will achieve sobriety.  The chart uses the individual values, which resulted in a skewed distribution. The following chart shows a similar bivariate comparison between the response variable and the groupings of distinct treatment type counts.

The relationship is similar to the ungrouped chart, though the proportion of sober respondents in the high treatment counts categories is lower -- as would be expected given the spread of values among the 9, 11, 12, 13 & 14 values.

 Both bivariate charts suggest what appears to be a strong positive correlation between number of distinct treatment types and achieving sobriety.

Comments

Post a Comment

Popular posts from this blog

First Program

 Assignment for Week 2: First Program Code : # -*- coding: utf-8 -*- """ Created on Sat Jul  6 10:48:29 2019 @author: gdeal """ # program to import nesarc dataset limited to respondents who ever needed to drink more to get the intended effect import pandas import numpy # option to avoid runtime errors/warnings pandas.set_option('display.float_format',lambda x: '%f'%x) # read in full nesarc dataset nesarc_df=pandas.read_csv('nesarc_pds.csv',low_memory=False) # convert variable names to upper case nesarc_df.columns=map(str.upper,nesarc_df.columns) # show number of observations and columns in base dataset # print(len(nesarc_df)) # print(len(nesarc_df.columns)) # print("printing values of sex for subset S2BQ1A2") # count_test = nesarc_df['S2BQ1A2'].value_counts(sort=False) # print(count_test) nesarc_df['S2BQ1A2'] = nesarc_df['S2BQ1A2'].convert_objects(convert_numeric=True) nesarc_df['AGE']
Read more

Program for Making Data Management Decisions

The following program creates a subset of the NESARC dataset that limits respondents to those who have had episodes in which they continued to drink to feel an effect and who were between 18 and 39 years old at the time of the survey. This subset is then used to identify respondents who have sought and received treatment for alcohol dependence. The variables include: Age & Sex of respondents Responses to questions on the type of treatment sought/received A variable was created to sum the number of different types of treatments respondents sought. Another variable was derived from that to indicate whether a respondent had received any type or treatment or not. Response values that did not indicate a yes or no to a question on treatment type were converted to NaN. Frequency counts and percentage distributions were computed for: Age Age Group Sex Number of treatment types sought Categories of number of treatment types sought Whether treatment of any kind was sought Thi
Read more