Part 1. Graphic Inquisition for Figure 1 (Lee et al.

, 2020)

A. Gestalt principles & visual structure (Todorovic, 2008)

The data graph presented is good for its purpose, but it violates the principle of proximity
because the gray lines are so packed together which makes it very hard to focus on the blue dots
(representing patients in this case). It also violates the Gestalt principle of simplicity, making the
graph much more complicated and harder to read than it needs to be. The multitude of elements
(blue dots, green mean line, light blue confidence interval) could make the data difficult to
differentiate for some readers. Simple shapes would make it easier to understand the main
message.

B. Keep it simple: Decoding & Operations

This graph has way too much information gathered (N, mean, regression, and confidence
intervals), an overload of details that makes it hard for the reader to comprehend in the way this
graph is designed.

The layers don’t help either, as mentioned above, everything overlaps (even with the legend).
It would’ve helped if the legend was horizontal instead.

C. Less is more: Chart junk & data-ink ration

The confidence interval and the N would’ve been enough to showcase. The individual
regression trends are interesting, but they could be shown in a separate graph. The data-ink ratio
is low – most of the elements hide the actual essential details (gray dotted line hides the N).

D. Graphical data integrity & lie factors

The graphical integrity is good in this example, there is no misleading data – good scales
and good intervals. On the other hand, because of the chaotic nature of its design, it could
influence the reader to look at the individual data, instead of focusing on the overall results.

E. Annotation & stand-alone readability

The legend is brief and squeezed together in a corner, also overlapped by data. On the other
hand, annotation is good, and it makes it easier for the reader to understand the graph.
Figure 1.

Distribution of activity scores over admission for the whole sample (n = 23).

Note. Reprinted from "Monitoring and predicting psychiatric patient progress using machine learning
and wearable sensor data," by H. Lee, N. N. Ta, N. T. Quyen, J. H. Kim, & M. Kwak, 2020, Scientific
Reports, 10, 17412 (https://doi.org/10.1038/s41598-020-74425-x). © 2020 by the authors.

Reference list

Lee, H., Ta, N. N., Quyen, N. T., Kim, J. H., & Kwak, M. (2020). Monitoring and
predicting psychiatric patient progress using machine learning and wearable sensor data.
Scientific Reports, 10, 17412. https://doi.org/10.1038/s41598-020-74425-x

Todorovic, D. (2008). Gestalt principles. Scholarpedia, 3(12), 5345.

Part 2. Graphic design component

Crime Rates in US States with Names Starting with the letter 'M' Based on Urban Population
(High vs. Low Urban Population)

Figure 1. Crime Rates in US States with Names Starting with the letter 'M' Based on Urban Population
(High vs. Low Urban Population)

This figure compares the crime rates (for assault, murder, and rape) per 100,000 residents
in US states starting with 'M.' The left panel shows states with high urban populations, while the
right panel shows states with low urban populations. Maine and Maryland exhibit the highest
rates of assault, while Mississippi shows relatively high murder rates compared to other states.
Crime rates in low/high urban populations in the states beginning with the letter “M” in
the United States (Figure 2)
Objective
The presented figure describes the crime rates in the US for states whose names begin
with the letter “M”. The type of crimes it covers are Murder, Assault, and Rape, that are
showcased in both low and high urban populations. The dataset used presents statistics per
100.000 residents in each state in the United States.
The purpose of the figure is to see if there is a pattern in the chosen states based on the
urban population level. By splitting the data into two panels (High Urban Population and Low
Urban Population), the figure showcases the comparisons side by side. The justification for
choosing only states beginning with the letter “M” has to do only with illustrative reasons:
including all states would overly complicate the investigation. This can also be seen as a random
but representative sampling. The chosen variables are Murder rate per 100,000 residents, assault
rate per 100,000 residents, and Rape rate per 100,000 residents.

Design choices
First, the chosen plot was a bar plot (geom_bar), as it made visualization easier for the
investigated variables across different states, focusing on the number of cases.
Second, the facets were split into panels (facet_wrap) to make easy comparisons on how
urbanization may relate to crime rates.
Third, the color scheme was changed from the initial version – the default option of
ggplot to attribute different colors had an unfortunate result of showing bright and cheerful
colors. This had to be changed (with scale_fill_manual) to align with the severity of the context it
explores and to represent it better.
Finally, the bars were distributed side by side with the dodge positioning (position =
“dodge”) option to avoid possible overlapping.

Interpretation
The figure shows that states with higher urban population show generally higher Assault
rates compared to the low urban population ones which could indicate a potential correlation
between the 2 variables. The Murder and Rape rates, however, seem to be more evenly
distributed and not as influenced by the level of urbanization.