Student

Student & College Information Analysis
using SpotFire®

Min-Ho Shin
Application Reports, CMSC838B,
Feb 29, 2001

Introduction

Information visualization gives us an intuitive and fast way of understanding data and making decisions from data. It can also lead us to some unknown but important aspects of data.
Using a dataset about student and college data, I would like to show you how SpotFire can help you to have an insight on this big data with ease.

DataSet Description

What About

Various Information about College & College Students for the 1993-1994 school year

Who Made

U.S. News & World Report

Where Came From

StatLib at Carnegie Mellon University ( http://lib.stat.cmu.edu/datasets/colleges/ )
Taken from 1995 U.S. News & World Report's Guide to America's Best Colleges
Used in American Statistical Association's 1995 Data Analysis Exposition.
The purpose of the Exposition is to encourage statisticians to demonstrate techniques, especially graphical, for analyzing data and displaying the results of an analysis

Data Size

1298 Schools

Description

This data is about SAT scores of new students, Application statistics of that year, cost information, graduation rates and severel additional informations such as the state where the school is located, pct of donators to alumni so on.

Variables

College Name	Applicant Enrolled	Book Cost
State	# of students of Top 10%	Personal Spending
Y*	Full Time Undergraduate Students	Pct of Faculty with Ph.D's
X*	Part Time Undergraduate Students	Pct of Faculty with terminal degree
Public(1) / Private(2)	In-state Tuition	Student/ Faculty Ratio
SAT(math)	Out-state Tuition	Pct of Alumni who donate
SAT(verb)	RoomBoard Cost	Expenditure per student
SAT(combined)	Room Cost	Graduation Rate
Application Received	Board Cost	Enrollment Rate*
Application Accepted	Additional Cost	Admission Rate*
* these fields are added by myself for analysis purpos only Y = vertical position of the state from 1 to 5 toward north X = horizontal position of the state from 1 to 6 toward east ( 6 is for Alaska and Hawaii ) Enrollment Rate = 100 * Applicant Enrolled / Application Accepted Admission Rate = 100 * Application Accepted / Application Received

Goal of Analysis

INTRODUCTION
	If you were a High School student preparing to get into university after graduation, this dataset would be the most useful one at the moment. You may want to know which school is the best for your academic ability as well as economic situations. Or you may question the possibility for you to get admissions from the schools you have in mind. However, looking at the table with over one thousand schools in it, you may get lost among the data. At this point, Information Visualization tools can help you. I'm going to demonstrate how you can use SpotFire® to explore the school data visually which means to understand intuitively.
QUESTIONS / PURPOSES
	Admission Rate
		Which variables the admission rate has correlations with ? ( Ability of students, Tuitions so on ) Which schools are the most competitive? Where are those competitive schools located in the following analyses ?
	Enrollment Rate
		Which variables the enrollment rate has correlation with?
	Graduation Rate
		Which variables the graduation rate has correlations with ? ( Tuition, Donators, Location of the school so on ) Which schools are easy to graduate from ?

Analyses by Visualization

Admission Rate & Top 10pct of Private Schools


This is the graph that shows the relation between the number of students in top 10 pct of SAT and admission rate for only private schools. As you see, they have a strong relationship. The group of schools on the right bottom with yellow color can be considered most competitive schools to get into. Names of those schools are shown in the table right beside of the graph to the right. And on the top of the graph, those with yellow color are schools that have 100% admission rate, which might be considered to be easiest school to get into.

Admission Rate & Top 10pct of Public Schools


This graph is same with previous one except this is for public schools. You can see similar pattern here. The schools at the right side depicts the schools where most competitive students are enrolled. All of those are University of Callifornia's except Georgia Tech. There are two more groups with lowest and highest admission rate. You can also find where the Univ. of Maryland at College Park is located in this graph in the middle of the graph.

* I created variable X so that it describes the horizontal position of the state from 1 to 6 toward east.

X	1	2	3	4	5	6
STATES	WA, OR, CA, NV	ID, MT, WY, UT, CO, AZ, NM	ND, MN, SD, IA, NE, KS, MO, OK, AR, TX, LA	MI, WI, IL, IN, OH, WV, KY, TN, MS, AL, GA, FL	ME, NH, MA, VT, RI, CT, NJ, DE, MD, PA, NY, VA, NC, DC, SC	AK, HI

EnrollRate of public schools in terms of location

This grapho shows the enrollment rates in terms of horizontal location of schools. as you see, schools at coastal area ( yello ) seems to have lower enrollment rates than non-coastal area.

Using 3D graph, it's more clear because the density can be perceived more easily.

Graduation Rate with Tuition

This depics how graduation rate increase as the amount of tuition become high. And at the right upper corner, you can see yello groups of highly competitive private schools which means they are very expensive but with high graduation rate.

Graduation Rate and Admission Rate

Lastly, I tried to find any relationship between graduation rate and admission rate. But the corelation is not so clear. So I eliminated private schools and devided public schools into two groups by coastal and noncoastal schools. You can see them below.

Graduation Rate and Admisson Rate of not Coastal public schools

This is graph for non-coastal public schools. You can't find any relation between admission rate and graduation rate.

Graduation Rate and Admisson Rate of Coastal public schools

But for public schools at coastal area, the relationship is much more clear than non-coastal area, which is interesting.

Critique

Interface is very intuitive
Screen arrangement is neat
Full text search for one column is very useful for locating data
no aggregation - count, sum or other statistical aggregation function would helpful
no statistical measure - some support for statistical measures like regression, co-efficiency analysis would help decision-making. But it exists now
in 2D, no way to express density - when data are overlapped, I can't figure out how many data are hidden except using 3D graph.
setting properties only applies for current visualization - like colors, shapes, labeling... they applies current visualization window
Full text search should be more powerful search features such as regular expression