USER GUIDE
BY
FIRDAUS GHAZALI
VICTORIA NEO
LI JIAYI
1. APP OVERVIEW
This RShiny App offers 5 different types of analysis, which can be conveniently accessed and
toggled from the navigation bar:
● Overview
● Multivariate
● Time Series
● Geo-network
● Panel Data
For all dashboard pages, it consists of three sections:
1. The navigation bar provides quick links to navigate between different types of analysis.
2. The Parameters Tuning Panel allows users to adjust input variables, statistical variables,
or even model tuning parameters without writing a single line of code.
3. This is the main plotting panel of the dashboard, which provides essential plots for EDA
and CDA, modeling results for different models. Some insights are also highlighted
within this panel.
The input panel (panel 2) and output panel (plotting panel) are highly interactive, with outputs
changing dynamically based on the input. Additionally, within panel 2, it is interactive among the
plots; clicking one plot highlights corresponding points of interest in the other plots. For detailed
interactivity, please refer to section 2.
Special Features: This app also provides a theme selection option, highlighted at point 4,
allowing users to choose between different themes of interest.
2. DETAILED GUIDE ON INDIVIDUAL PAGE
Overview
This section provides an overview of the
project whereby the user is able to see some
of the key metrics as well as to conduct
simple EDA.
Panel 1 provides some key insights of the
dataset
Panel 2 allows the user to choose from
1. different input such as country,
variables of interest, and year
2. different statical testing method
Panel 3 consists of 3 key plots for
visualizations. It provides 2 different
interactivity.
1. These visualizations will be updated
automatically based on the input, and
the interest point will be highlighted in
green.
2. the user can click on any point of
interest within the panel, the
corresponding point of interest will be
highlighted between the plots
Multivariate
This section of analysis provides an overview
of analysis among different variables. It is
cross sectional, and focuses on a particular
year only.
Panel 1 allows the user to choose from
different year of interest, the statistical
argument was not exposed due to
1. The limitation of the library chosen.
The team has decided to go for
interactivity
2. There is insignificant difference upon
offline testing between different
statistical method due to the nature of
the dataset
Panel 2a will show up and get updated
immediately with the change of user input in
Panel 1.
Panel 2b corresponding scatter plot will show
up when the user click on any pixel in 2a
Panel 1 allows the user to choose from
1. different year of interest
2. Fine tuning of clustering models
including variables such as: data
transformation method, number of
clusters for rows/columns, linkage
methods, distance methods etc.
3. Variables of interest for clustering
results visualization
Panel 2 will only show up after model
buildings. Left side (2a) shows the clustering
results, right side (2b & 2c) allows users to
analyze and visualize the clustering results
Panel 2a shows the clustering results with a
dendrogram upon model buildings and gets
updated with different model calibration. The
user can hover to any of the links and pixels
to understand the particular data.
Panel 2b is consolidated data from clustering.
User can sort for different columns
Panel 2c will get updated when the user
change the interest of variables
Time Series
This section of analysis provides an overview
of analysis across different time periods for
Big mac index related data across countries.
Panel 1 allows the user to choose from
1. different variables of interest
2. Different aggregation techniques
Panel 2 is a horizon plots get updated based
on the user input
Panel 1 allows the user to choose from
1. different variables of interest
2. Fine tuning of clustering models
including variables such as: clustering
method, number of clusters for
rows/columns, linkage methods,
distance methods etc.
3. Variables of interest for clustering
results visualization
The panel consists a few storyboards
whereby the first board focus on model
tuning and selection, the rest for visualization
Panel 2a is by default selected for user to test
a range of cluster numbers
Panel 2b shows the results of various k
variables and the evaluation scores are
displayed in a table
Panel 2c provides a guide on how user
should choose and evaluate the performance
whereby allows the user to decide on 1
optimal k value for model building
Panel 3 provides a variation of visualizations
for the user to visualize and analyze the
clustering results.
The user can select between panels at 3a,
and the corresponding plots are updated at
3b.
Geo-Network
This section of analysis provides an overview
of network and geospatial analysis across for
Big mac index related data across countries
in 2021. It focuses on trade networks on beef.
This page is the EDA for geo-network
analysis.
Panel 1 allows the user to control the layouts
preferred for Panel 2.
Panel 2 consists of network visualisations
that will change according to the User’s
selection in Panel 1.
Panel 3 is a visual network graph of
connections between countries. User can
select the country in the box on the top left
and see its connections being highlighted.
Panel 1 allows the user to select the centrality
measure of interest. This will control the
centrality results in Panel 2, 3 and 4.
Panel 2 shows the visualization of centrality
analysis and this result is visualised as a bar
chart in Panel 3. The results are captured for
exploration in Panel 4. All types of centrality
measures are patched together in Panel 5.
Panel 1 allows the user to adjust the
parameters for both the Walktrap and
Spinglass models and these are visualised in
Panels 2 and 5.
If users are unaware of which steps (for
walktrap) or gamma (for spinglass), they can
expand Panels 3 and 6 to find the optimal
values. The clustering results are captured in
Panels 4 and 7 respectively.
Panel 8 shows the modularity of each model
and the comparison between the two.
Panel Data
This section of analysis provides an overview
of panel data analysis across different time
and country dimensions for Big mac index.
Panel 1 allows the user to choose their
preferred variables as inputs to the plots.
Panel 2 is an animated bubble plot. User can
click Play to see how the data points change
through time. The selected country from
Panel 1 is green in colour.
Panel 3 is a coplot of the selected variable
against year through different countries.
This page is in a storyboard layout.
Panel 1 are the titles of the different frames of
the storyboard. User can scroll right to see
the progression of the story.
Panel 2 is the main plot of the story frame.
Panel 3 is the description of the main plot.
Panel 1 allows the user to choose preferred
variables and selected parameters to build
either a ols or panel data regression model.
Panel 2 displays the results of the regression
model built.
Panel 3 displays the estimated-observed
values plot of the regression model built.
Interested to know more?
Please visit our website @ https://bmi-pricebite.netlify.app/
