Week 4 - Creating Graphs for Data

The data from the Gapminder data set consisted of quantitative variables. Thus, the appropriate methods were used to analyse the relationships between the data.

Univariate Graphs

Univariate graphs were generated for each of the 3 variables - co2emissions, relectricperperson and urbanrate. All of the variables were quantitative and therefore histograms were generated to represent the distribution of the data. The program used is shown below, as well as the output graphs.

The Program


Univariate Graph for co2emissions

This graph is unimodal with most observations having cumulative emissions around 0.1e11 or 1,000,000,000 (i.e. 1 billion)  metric tons. The graph is also skewed right with most observations having the lowest emissions. There is also an outlier at round 3.4e11(34 billion) metric tons.

Univariate Graph for relectricperperson

This graph also seems to be unimodal and skewed right, with the peak at around 250 kWh and most observations having the lowest recorded consumptions.

Univariate Graphs for urbanrate

This graph also shows a unimodal distribution with a peak at around 75% urban population. This is the only distribution that is skewed left with most of the observations having higher percentages of urbanisation.

Bivariate Graphs

Three bivariate graphs were generated to analyse the relationships between the chosen variables. Depending on the association being analysed, the variables were classed as either explanatory (E) or response (R) as follows:
  1. Urban rate (E) and CO2 emissions (R)
  2. Urban rate (E) and residential electricity consumption (R)
  3. Residential electricity consumption (E) and CO2 emissions (R)

1. Urban rate (E) and CO2 emissions (R)

The Program


The Graphs

In this graph, the majority of the points are clumped near the bottom because nearly all of the observations had very close values for CO2 emissions. This made analysis difficult, therefore a bar graph was generated based on quartiles of urbanisation.

The bar graph showed that countries with higher urban rates (i.e. those in the higher percentiles) had larger mean cumulative CO2 emissions. Therefore, based on the data, there is a positive relationship between these two variables.

2. Urban rate (E) and residential electricity consumption (R)

The Program


The Graph


The graph of electricity consumption vs % urban population shows what seems to be a positive relationship between the two variables. As the % urban population increases, so does the electricity consumption.

3. Residential electricity consumption (E) and CO2 emissions (R)

The Program


The Graphs

Similar to the graph of co2 emissions vs urban rate, the points on this plot are clustered near the bottom of the graph, indicating a need to generate a bar chart for deeper analysis. 

The bar chart shows that generally, as electricity consumption increased, so did CO2 emissions. This implies a positive relationshp between the two variables.



Comments

Post a Comment

Popular posts from this blog

Week 1 - Getting the Research Project Started

Codebooks and Datasets Selecting the Variables As I am interested in pursuing a career in renewable energy and sustainability, one of my main areas of interest was carbon dioxide (CO2) emissions. From  my studies, I know that some fuels used in the generation of electricity produce these emissions and therefore, I chose the following two variables from the Gapminder codebook: co2emissions - the cumulative metric tons of co2 produced from 1751 to 2006 relectricperperson - the residential electricity consumption per person during 2008 in kWh Additionally, increased urbanisation implies increased human activity which means increased amounts of carbon dioxide emission. Thus a third variable was selected from the dataset:      3. urbanrate - percentage of total population living in urban areas in 2008 Research Question  Are CO2 emissions associated with the levels of residential electricity consumption and urbanisation? Hypothes...
Read more

Week 3 - Making Data Decisions

Image
For this assignment, due to the nature of the dataset chosen, the only data management that was necessary was coding out missing data. In fact, as I discovered,  the software had already disregarded the missing data so this was not an absolutely necessary step - however, it was a learning moment and allowed me to verify what I had assumed in last week's assignment. I.e. The total number of observations (countries) was 213 based on the length, or number of rows, of the dataset. In the following I will discuss how I determined if this assumption was correct or not. Furthermore, I had unknowingly already executed the technique of grouping variables with pd.cut in last week's assignment so that I could demonstrate and use frequency tables. Replacing the Blanks with 'Nan' This was done as in the video with the following code (program variable for co2 used as the example): c1= c1.replace(r'^\s*$', np.nan, regex=True) where c1 was the program variable assigne...
Read more

Week 2 - Running My First Program

Image
I chose to use Python software to perform the data analysis. The data from the Gapminder data set was categorical so carrying out the frequency programming as shown in the video gave results of 1 for each count as well as the same percentage value for each variable. An example of this is shown in the following 2 images. This was because the values for the chosen variables were unique to each country. After looking through the discussion forum, I decided to try one of the suggested solutions which was to use pd.cut to create 'bucket categories'  for each variable. The code shown below was used to create 4 intervals of the values of the CO2 emissions variable, and label each interval. co2 = pd.cut(mydata["co2emissions"],4,labels=['low emissions','low-mid emissions','high-mid emissions','high emissions']) This was done for each variable and then I adjusted the code to work with the newly created grouped variables ( ...
Read more