Data preparation and cleaning for my datasets entailed merging the
reformatting the datasets and combining them. To prepare the CHR&R
dataset for data merge, the most relevant variables were selected for
analysis and all other remaining variables were dropped. Because the
data is organized by county and the state level data is needed, the
county level data was aggregated by state. There was not much that
needed to be done to prepare the SAMHSA dataset for merge, since it
was already properly formatted. The two datasets were merged using
R which facilitates joining two data frames. In this case, a left join was
used to join the SAMHSA (on the left) and CHR&R (on the right)
datasets based on state. This means that the final output table
contains all the rows and columns of data in the SAMHSA dataset and
the corresponding matching column data in the CHR&R dataset. Data
cleaning including handling missing data and removing outliers, all of
which was done using R.
The analytical tools used in my project are Tableau, R and Python.
Tableau is a great user-friendly tool for creating visualizations of
geographic data in the form of data maps. The drawback to using
Tableau is that it does not offer users many customization options.
Overall, however, Tableau’s click-and-drag functionality make it a
good choice for creating quick and informative maps. Mapped
visualizations are important in this exploration because the mental
health landscape will be compared across states. Initially, it will be
important to explore various health and environmental parameters by
state to gain a better understanding of the data.
R is also great tool choice for visualizations and is used in this project
for data preparation, cleaning, and initial data exploration. R was
specifically built for statistical analysis and is particularly well known
for generating some of the most attractive data visualizations (IBM,
2021). In R it is easy to generate statistical summaries in one line of
code by using simple commands like str() and summary(). It is also easy
to create visualizations for univariate and multivariate analysis of each
field in the dataset. It is also capable of handling large datasets.
Python is the tool of choice for the predictive model development
since it is a general-purpose tool widely used for big data analysis and
machine learning. As of 2022, it is considered the top programming
languages by popularity (Luna, 2022). Although R is also useful for
modeling large datasets, Python is more intuitive to use since the code
is more readable. Since Python is open source, it will also be easy to
find solutions to any potential issues through online forums, analytics
libraries, and other online documentation. Python will make it is easy
to model the data and generate easy to interpret summaries of the
results. Several machine learning methods will be employed in this
project and the performance of each will be compared. Since the
target variable is categorical, neural networks, SVM, random forest,
and ensemble models will be used which can all be implemented using
Python.
Sources:
IBM Cloud Team & IBM Cloud (2021).
Python vs. R: what’s the
difference?
IBM. https://www.ibm.com/cloud/blog/python-vs-r
Luna, J. (2022). Top programming languages for data scientists in
2022. Datacamp. https://www.datacamp.com/blog/top-
programming-languages-for-data-scientists-in-2022
