|
Çağlar
Aksezer
|
email:
|
caksezer@coe.neu.edu
|
|
|
research/project:
|
Multivariate
Efficiency and Optimization Problems in Quality Engineering
|
|
joined
QPL:
|
September
2000
|
|
hometown:
|
Istanbul,
Turkey
|
|
education: |
BS,
Mathematical Engineering, Yildiz Technical University (1999)
MS,
Industrial Engineering, Northeastern University (2000)
|
|
where
is he now: |
Boston,
MA |
|
Description
of Work: Caglar’s
research is addressing two general problems that arise in quality
engineering when attempting to evaluate or optimize processes with
multiple responses, outputs, or inputs.
The first area examines the use of multivariate quadratic loss
functions to determine optimal process parameters and compares these
results to non-linear and desirability function programming approaches.
General expressions have been derived for the variances and
probability density functions of the three most common loss functions
(nominal-the-best, smaller-is-better, larger-is-better) and simplified for
cases where the response is normal, lognormal, Weibull, exponential, and
uniform. In many cases, the
results exhibit significant variance, skewness, and unique asymmetrical
shapes. Several mathematical
programming formulations have been developed based on these results that
minimize the total loss variance or expectation, possibly subject to
probabilistic constraints on individual loss.
These models currently are being tested and illustrated on several
single- and multiple-response examples.
A second research area concerns the effect of missing data in data
envelopment analysis (DEA) frontier estimation, a common problem
encountered in practice. Caglar
is comparing the results of several possible approaches for various
amounts of missing data (0%, 1%, 5%, 10%, etc), including variable
deletion, best and worst case replacement, bootstrapping, and multiple
imputation methods, using a library of twenty-four benchmark DEA problems
taken from the literature. These results will be useful to DEA
practitioners and researchers when faced with incomplete data sets.
|
|
Hobbies:
Technical diving, nautical archeology
|
|
|
|
Kerri
Beiswenger
|
email:
|
beiswenger.k@neu.edu
|
|
|
research/project:
|
Lab
Undergraduate Assistant. Appointment
Access Across Organizational Boundaries
|
|
joined
QPL:
|
April
2003
|
|
hometown:
|
New Hartford, NY
|
|
education:
|
BS
candidate (2005), Industrial Engineering,
Northeastern University
|
|
where
is she now:
|
Intel,
Hudson, MA
|
|
Description
of Work: Kerri
has been helping with several ongoing projects in the QPL lab, as well as
assisting with data analysis and literature reviews.
Kerri also has been working on a large multi-facility project to
study patient appointment access, flow problems, and prolonged delays in
referral processes and across organizational boundaries. This project may include the development of queueing
network models and simulation decision support tools to help redesign
inter-facility access and to better understand the current process.
|
|
Hobbies:
Running, hiking, water skiing, photography
|
|
|
Alicia
Borgman
|
email:
|
aborgman@coe.neu.edu
|
|
|
research/project:
|
Statistical
Process Control Methods for Non-homogeneous Dichotomous Processes
|
|
joined
QPL:
|
September
2001
|
|
hometown:
|
Pleasant
Grove, CA
|
|
education:
|
BS,
Applied Mathematics, California State University (2001),
Candidate
for MS, Operations Research, Northeastern University
|
|
where
is she now:
|
Boston,
MA
|
|
Description
of Work: Alicia
is investigating methods for applying statistical process control (SPC) to
non-homogeneous dichotomous Bernoulli data.
Important applications include defective items produced by
different processes, automobile accidents aggregated across driver types,
and surgical site infections in patients with different risk factors. A new mixed-risk probability distribution, called the J-binomial,
and associated control charts have been developed, and their properties
(moments, average run lengths, etc) currently are being investigated and
compared to those of corresponding binomial, normal, and other
approximations. These
distributions also have been compared using a modified Kullback-Leibler
information statistic, a total absolute deviation metric, a variance
ratio, and experimental design methods.
The J-binomial random variable also has been proven mathematically to
be under-dispersed with respect to its binomial counterpart.
Appropriate Shewhart, EWMA, and standardized control charts have
been developed and investigated using both k-sigma
and probability limits. A
special-purpose software program has been developed to aid practitioners
and analysts with the computational complexity of this new model, which
requires a series of J-1 nested recursive convolutions, and the calculation of
probability and cumulative distributions, control limits, and average run
lengths. All results are
being verified via Monte Carlo simulation and applied to several empirical
healthcare and manufacturing data sets.
|
|
Hobbies:
Snowboarding,
photography, swing dancing, sewing
|
|
|
Adrienne
Fusco
|
email:
|
adfusco@coe.neu.edu
|
|
|
research/project:
|
Undergraduate Research Assistant: Statistical
Models of Patient Safety
|
|
joined
QPL:
|
September
2002
|
|
hometown:
|
Queensbury,
NY
|
|
education:
|
BS,
Industrial Engineering, Northeastern University (2003)
|
|
where
is she now:
|
Lehigh
University, Bethlehem,
PA
|
|
Description
of Work:
Adrienne has worked as a QPL coop student and undergraduate research
assistant on several ongoing projects.
A large part of her focus has been on developing a Monte Carlo
simulation program to investigate and compare the performance of new
negative binomial exponentially weighted moving average (EWMA) and
cumulative sum (Cusum) statistical process control methods with their
binomial and Bernoulli counterparts.
These methods are being designed for certain health care and
patient safety problems of large concern across the U.S.
(The national costs of medical errors and other adverse events are
staggering, including an estimated 770,000 to 2 million injured patients,
44,000 to 180,000 deaths, and $8.8 billion additional healthcare costs
annually.) Adrienne also has assisted with the application of these
methods to several large hospital data sets.
Results of this work are being incorporated into a working paper
being prepared for journal submission.
Adrienne’s educational objectives are to gain a better
understanding of statistical monitoring methods, to gain insight into the
scope and conduct of this sort of research project in general, and to help
explore possible interest in pursuing graduate studies.
|
|
Hobbies:
Reading, movies, visiting friends
|
|
|
Alp
Gumus
|
email:
|
alpgumus@coe.neu.edu
|
|
|
research/project:
|
Optimal
Dual Control Chart Monitoring Schemes for Bernoulli Processes
|
|
joined
QPL:
|
September
1999
|
|
hometown:
|
Istanbul,
Turkey
|
|
education:
|
BS,
Bogazici University (1999)
MS,
Operations Research, Northeastern University (2001)
|
|
where
is he now:
|
Verizon
Data Services
|
|
Description
of Work: Alp's
research is focusing on the performance and optimal design of combined
attribute control chart schemes based on combinations of p (number-within)
and g (number-between) Shewhart, EWMA, and Cusum control charts.
Alp is developing numeric and Monte Carlo programs to investigate
the performance of these schemes. Preliminary
results indicate that using more than one chart of either different types
or with different subgroup sizes can significantly increase the ability to
detect process shifts in either direction, a noticeable limitation of many
single attribute control charts. Alp
previously developed Java program to investigate the performance and
optimal economic design of single monitoring approaches, and now also is
developing economic and optimization models to explore the optimal design
of any given dual monitoring scheme.
|
|
Hobbies:
playing
chess, watching movies & documentaries, travel, outdoor activities,
tennis
|
|
|
Amit
Nene
|
email:
|
NeneA@chesterton.com
|
|
|
research/project:
|
Performance
and Optimal Subgroup Sizes for Number-Between g Control Charts
|
|
joined
QPL:
|
September
2000
|
|
hometown:
|
Mumbai,
India
|
|
education:
|
B.S.
Mechanical Engineering, Mumbai University (1999)
M.S.
Industrial Engineering, Northeastern University (2002)
|
|
where
is he now:
|
AW
Chesterton, Stoneham, MA
|
|
Description
of Work: Amit's
research is focusing on the performance and optimal design of
number-between g- and h-type statistical control charts.
These charts are based on inverse sampling from negative binomial
distributions, rather than the conventional np, p, EWMA, and related
charts based on (positive) binomial distributions.
Several numeric programs have been developed to calculate the exact
two-sided and one-sided probability limits and their associated expected
run length (ARL), expected number of items (ANI), and standard deviations
until the chart detects a process shift.
The performance of these charts has been exhaustively explored
under a variety of conditions and for a wide range of subgroup sizes.
Amit also has been investigating the optimal subgroup size that
minimizes the expected number of items until detection of a process shift
and comparing these results to traditional binomial Shewhart and EWMA
charts. Preliminary results
are counter-intuitive and suggest that a subgroup size of n = 1 (e.g.,
number until the first defect) is rarely optimal.
|
|
Hobbies:
sports - badminton, cricket, table tennis,
tennis and more; tracking international news and affairs, trekking
|
|
|
Stephanie
Mason
|
email:
|
smason@coe.neu.edu
|
|
|
research/project:
|
Undergraduate Research Assistant: Application
of New SPC Methods to Patient Safety
|
|
joined
QPL:
|
April
2002
|
|
hometown:
|
Newport
News, VA
|
|
education:
|
BS,
Industrial Engineering, Northeastern University (2003)
|
|
where
is she now:
|
DSC
Logistics, Montgomery, IL
|
|
Description
of Work: Stephanie
is working as an undergraduate research assistant on several ongoing
projects, as well as heading up our website and tools development efforts.
A large part of her focus has been on analysis of healthcare safety
and infection control data using special-purpose statistical process
control (SPC) methods developed by the QPL (for rare event and
non-homogeneity problems). Important applications of the first type include infrequent
but catastrophic employee accidents, patient medication errors, and
manufacturing defects. Important
applications of the second type include different likelihoods of accidents
between facilities, infections patient-to-patient, and defects between
product types. The National
Academy of Sciences recently estimated that more Americans die each year
from adverse events while patients in U.S. hospitals than from traffic
accidents, breast cancer, or AIDS, with roughly $9 billion spent annually
as a result of medical mistakes. Stephanie
also has been developing spreadsheet templates for the above methods that
are available on the Tools (LINK) section of this site.
|
|
Hobbies:
Singing, traveling, relaxing
|
|
|
|
Arianne
Nartyasari
|
email:
|
annecantik@yahoo.com
|
|
|
research/project:
|
Optimal
Design of Healthcare Statistical Process Control Charts
|
|
joined
QPL:
|
January
2001
|
|
hometown:
|
Taiwan
|
|
education:
|
BS,
Industrial Engineering,
Northeastern University
|
|
where
is she now:
|
Teradyne
(North Reading MA)
|
|
Description
of Work: Arianne’s
research concerns the optimal design of statistical process control (SPC)
charts for healthcare processes. While
control charts are being used with increasing frequency in many healthcare
applications, there continues to be some debate as to how differences
between healthcare and more traditional manufacturing applications should
affect the appropriate subgroup size, control limits, and resulting
performance. Important
healthcare applications of SPC include infection control, patient
management, disease surveillance, surgical site infections, medication
errors, and other adverse events – all process “defects” with
possibly different consequences than in more traditional applications.
Arianne therefore is applying mathematical control chart economic
optimization models to a wide range of healthcare applications in order to
explore the results in these settings.
These models determine the optimal subgroup size, sampling
interval, and control limit width for any given set of inputs, and have
been implemented in several computer programs.
Results of this study are being summarized and compared to current
practices and more traditional applications, along with extensive
sensitivity analysis on inputs and assumptions.
This work also will provide useful guidelines for practitioners
using SPC throughout healthcare.
|
|
Hobbies:
|
|
|
Ipek
Ozer
|
email:
|
ipekozer@coe.neu.edu
|
|
|
research/project:
|
Bounded
Feedback Adjustment and Statistical Monitoring Methods for Autocorrelated
Healthcare Processes
|
|
joined
QPL:
|
January
2002
|
|
hometown:
|
İstanbul,
Turkey
|
|
education:
|
BS, Mechanical Engineering, İstanbul Techical
University (1999)
MS, Mechanical Engineering, Northeastern
University (2002)
|
|
where
is she now:
|
Boston,
MA
|
|
Description
of Work: İpek’s
research is integrating statistical process control (SPC) and
bounded feedback adjustment (BFA) methods for healthcare data that exhibit
autocorrelation or for which it is desirable to make periodic adjustments
to minimize deviations from a target value.
Important applications include disease incidence, respiratory
illness, seasonal data, oral anticoagulants, oxygen saturation levels in
ICU patients, and pituitary hormone regulation – where competing costs
of adjustments, deviations from desired levels, and delayed change
detection need to be balanced. As
an alternative to continuous feedback control, bounded control adjusts a
process only when a monitoring statistic of the time series falls outside
specified bounds, motivated by applications in which continual adjustments
are not practical or have significant costs.
İpek’s research
therefore is developing new process control methods in which
periodic feedback adjustments are used to minimize deviation and SPC is
used to detect changes in the underlying time series process.
This work includes identifying underlying time series models for
given applications, determining statistical properties of the bounded
adjustments, and evaluating the performance of alternate hybrid approaches
under different types of process shifts.
Cost models also are being developed to determine the optimal
simultaneous design of the two methods together and their robustness to
model misspecification.
|
|
Hobbies:
photography,
sailing, oil painting, biking
|
|
|
Ashwini Tumne
|
email:
|
tumne.a@neu.edu
|
|
|
research/project:
|
Special-Purpose
SPC Software
|
|
joined
QPL:
|
January
2002
|
|
hometown:
|
Bombay,
India
|
|
education:
|
BE,
Computer Engineering, University of Bombay (2000)
MS, Computer Science, Northeastern
University (2003)
|
|
where
is she now:
|
Boston, MA
|
|
Description
of Work:
Ashwini
is developing commercial-grade statistical process control (SPC) software
that constructs standard and special-purpose SPC methods developed by the
Quality and Productivity Laboratory (QPL), as well as providing general
programming and web support to various QPL projects.
The current alpha version of the SPC software is developed in an
object-oriented environment and includes for all standard Shewhart,
exponentially-weighted moving average (EWMA), cumulative sum (Cusum), and
standardized control charts, as well as several new methods for rare
events, short run, autocorrelated, and feedback adjustment processes.
Several modules are specifically tailored to healthcare
applications for monitoring such concerns as ventilator-associated
pneumonia, surgical site infections, medication errors, patient
physiologic variables, disease surveillance, respiratory illness, and
other seasonal data and adverse events.
A working beta version of the software should be available over the
QPL website in summer 2003.
Please contact us
for further information.
|
|
Hobbies:
Reading,
roaming, meditation
|
|
|