Development, Contribution and Application of Statistical Methodology
Health Tag:Dental Epidemiology, Infectious Disease, Chronic Disease, Drug Evalu-
ation, Case-Control Study, Competing risk model.
Statistics Methodology Tag:Count Data, Poisson, Negative Binomial, Missing
Values, EM algorithm, Mixed E.ect Modeling, Logistic Regression, Gamma Regres-
sion, Score Test, Likelihood Ratio Test, Markov Chain, Multi-State model, Copeting
risk survival, Prediction modelling.
Project Title: Prediction modelling using Multi-State Model for Multiple Decrement / Competing
Risk Survival Data.
Organization, Duration & Role: SickKids, 2017-present, Biostatistician.
- Project Synopsis: Finding best prediction model for analysing competing risk survival data
based on multistate model as well as competing risk model. Look for best possible way to check
the performance of prediction model by using receiver operating characteristics.
- Application: Simulating multistate data and building prediction model. Extracting transition
probabilities for decision modelling. Applying the methods to the administrative data from ICES
(Institute for Clinical Evaluative Sciences).
- Software: R (data management, preparation, and analysis; packages: msm, mstate).
Project Title: Zero Inflated Over Dispersed Count Data with Missing Values [Part of this project
has published in Statistics in Medicine].
Organization, Duration & Role: UWindsor, 2011-2016, Post-Doctoral fellow, PhD student.
- Project Synopsis: Regression analysis of zero in
ated over dispersed count data is complex and
it becomes further complicated by the existence of missing values either in the response variable
and/or in the explanatory variables. In this project, I have developed an estimation procedure
via weighted EM algorithm for zero in
ated over dispersed count data regression model at the
presence of missing response and /or covariates.
- Application: Estimation procedure have been applied to a dental epidemiology data from a
caries prevention study, famously known as decayed, missing and .lled teeth (DMFT) index data.
DMFT index is one of the most commonly used method to assess dental caries prevalence as well
as dental treatment. I have .tted the zero in
ated negative binomial model to this data. Direct
maximization and EM algorithm have used for the complete data, and the data with missing
values respectively.
- Software: SAS (data management, preparation, and analysis; proc: countreg, genmod, nlmixed),
R (simulation and model .tting; packages: maxLik, nleqslv, gamlss, VGAM, pscl, MCMCglmm,
glmmADMB).
Project Title: Two Part Random E.ect Models for Semicontinuous Data [This project has pub-
lished in International Journal of Statistics in Medical Research].
Organization, Duration & Role: UNB, 2009-2011, Research Assistant, MSc student.
- Project Synopsis: Toenail onychomycosis is a very common toenail disease. Lamisil and
itraconazole were two di.erent oral treatments for this disease. To compare the e.cacy and the
safety features of these two treatments, I applied the two part random e.ect models to the data.
I have used the mixed e.ect logistic regression model and the mixed e.ect gamma regression
model in the .rst and the second part of the models respectively.
- Application: The toenail data was collected through a randomized, multicenter, double-blind,
parallel-group study. The study was operated to evaluate the e.cacy of two antifungal com-
pounds (lamisil and itraconazole) that reduced the duration of treatments.
- Software: R (data manipulation and analysis; packages: MASS).
Project Title: Development of Test Statistics for some Parametric Over/Under-dispersed Life Time
Models [In press: Acta et Commentationes Universitatis Tartuensis de Mathematica]
Organization, Duration & Role: UWindsor, 2011-2015, Research Assistant, PhD student.
- Project Synopsis: To .t a model for waiting time in the emergency room, weekly rainfall or
river discharge volumes, exponential distribution can be an appropriate choice. This exponential
distribution is a special case of a more richer family of distributions, such as the Pareto distri-
bution, the gamma distribution and the Weibull distribution, all of which are two parameter
distributions and can be expressed as a family of over-dispersed exponential models. I have
developed Score and Likelihood Ratio Test (LRT) statistics to test the goodness of .t of the
exponential model against the over-dispersed family of distributions.
- Application: The score and the LRT statistics were applied to two di.erent engineering data:
the lifetime for the Kevlar/Epoxy strand at 70% stress level, and to the breakdown times of light
bulbs at di.erent levels of voltage.
- Software: R (data manipulation, simulation and analysis; packages: nleqslv, actuar, survival).
Project Title: Markov Chain Approach for analysing Diabetes Mellitus Data [This project has
published as a monograph from Germany].
Organization, Duration & Role: UnivDhaka, 2008-2009, Research Assistant, BSc and MSc student.
- Project Synopsis: Transition probability is very important for Markov chain. In this project, I
estimated the transition probability of Markov chain using maximum likelihood estimation. Or-
der test and the test of homogeneity or stationarity test was performed to identify the properties
of Markov Chain.
- Application: I was actively involved in primary data collection of a diabetic hospital. Data was
collected from diabetic patients logbook. I used a portion of that data to estimate the transition
probabilities of the Markov chain, and perform the order test, and the stationarity test to identify
the properties of the Markov chain.
- Software: R and SPSS (data management and analysis).
. Administrative Database Management and Data Analysis
Tag: CIHI (DAD), Primier, Case-Control Study.
Statistics Methodology Tag: Kaplan Meier, Cox-Proportional Hazard, Multinomial
Logistic Regression, T-test, Chi-Square Test, Log Rank Test.
Project Title: Patterns of Malignancy Following Transplant (using CIHI (DAD) data).
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Finding Pattern of malignancy after solid organ and stem cell transplant.
- Application: Complete exploratory and survival analysis, data linkage between Canadian Can-
cer Registry. Preparation of reports and possible manuscripts for scienti.c publications.
- Software: SAS and R.
Project Title: Intravenous immune globulin (IVIG) utilization as an adjunctive therapy in patients
with septic shock (Using Primier Database).
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: IVIG is known to be a plasma product derived from human serum and ad-
minister to the patients with immune de.ciency. Goal of this project is to evaluate the utilization
pattern and e.ectiveness of IVIG.
- Application: A propensity score matched retrospective cohort study was designed to capture
and organize data. We are working on multivariate mixed e.ect regression model and trend
analysis to explain e.ectiveness and utilization of IVIG for severe sepsis.
- Software: SAS and R (data management, visualization, analysis); Packages: SAS[Proc sql,
logistic, rank, freq, sort, univariate, lifetest, phreg, sgplot, gplot], R [MatchIT, dplyr, ggplot,
coxph, surv.t].
. Survey, Study Design and Screening
Tag: REDCap, CAISIS, Maxon, Database Management, System Performance, Pa-
tients Reported Outcome, Validation.
Project Title: Canadian Partnership for Tomorrow Project - Manitoba [Canadian Partnership
Against Cancer(CPAC) funded].
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Create and manage a database with REDCap or CAISIS- to collect, store
and link coded participant questionnaire data, in a secure, encrypted, password protected envi-
ronment on CCMB servers. Create a data dictionary for the questionnaire (using the national
example provided) that is compatible for harmonization with CPTP. Create linkages between the
questionnaire database and a key .le. The key .le - containing participants names, addresses,
PHIN, and study ID numbers- must be stored on a separate server from the questionnaire data.
- Application: Create and manage a database, data dictionary and linkages between databases.
- Software: REDCap (Research Electronic Data Capture), CAISIS (patient data management
system), SAS and R .
Project Title: Patients Reported Outcome Survey: AOPSS (Ambulatory Oncology Patient Satis-
faction Survey).
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Create and manage a database and perform full scale statistical analysis.
- Application: Tabulation, Association, system performance.
- Software: Excel, SAS and R .
Project Title: Patients Reported Outcome Survey: The Comprehensive Problem and Symptom
Screening (COMPASS) .
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Create and manage a database and perform full scale statistical analysis.
- Application: Tabulation, Association, system performance and survival.
- Software: Excel, SAS and R .
Observational Study: Cohort and Case Control Study
Health Tag: Genomic markers, Chronic Lymphocytic Leukemia (CLL), Tracheal
Cancer
Statistics Methodology Tag: Kaplan Meier, Competing Risk, Cox-Proportional
Hazard, Time varying Covariate, Logistic Regression, T-test, Chi-Square Test, Log
Rank Test.
Project Title: Population based clinical trial on Chronic Lymphocytic Leukemia (CLL): Statistical
analysis for Skin and Second Cancers cohort of 10 years in CancerCare Manitoba.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Over 600 CLL patients have been followed over 10 years period. This
cohort is still active. The intention of the study is to provide large scale statistical analysis for
all possible measures of the cohort.
- Application: Complete exploratory and survival analysis, identifying genomic markers and
model .tting for the CLL cohort. Preparation of reports and presenting them in biweekly
meeting as well as preparing possible manuscripts for scienti.c publications.
- Software: SAS and R.
Project Title: Clinical Characteristics and Prognosis of Primary Tracheal Cancer: A Single Institu-
tion Experience (Largest Canadian Series).
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Exploring di.erent characteristics of tracheal cancer and .nding overall
survival.
- Application: Complete exploratory and survival analysis. Preparation of poster and manuscript
for scienti.c publication.
- Software: SAS and R.
. Experimental Study: Randomized Control Trial
Tag: Sample size Calculation, Diagnostic Tests, Childhood Leukemia
Statistics Methodology Tag: Online Calculator(PASS), Concordance, Sensitivity,
Speci.city, Predictive Values (PPV, NPV), ROC curve, Prevalence, Likelihood Ratio,
Chi-Square Test.
Project Title: Sample size determination for a potential non-inferiority randomized trial small cell
lung cancer patients.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Discuss and provide statistical support for sample size determination and
feasibility of the study for non-inferiority randomized clinical trial.
- Application: Reviewing literature and translating scienti.c knowledge as well as providing
possible sample size to the Principle Investigator (PI) .
- Software: R and online calculator.
Project Title: Comparison of Light Transmission Aggregometry (LTA) and Multiple Electrode Ag-
gregometry (MEA) for the Evaluation of Platelet Function in Patients with Mucocutaneous Bleeding.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Comparison between two diagnostic methods (LTA and MEA) for evaluation
of blood platelet function. MEA is is relatively newer which require smaller sample size with less
manipulation. LTA is considered as standard methods for evaluating platelet function.
- Application: Evaluating diagnostic methodologies for for blood platelet function by using sta-
tistical methodology: Sensitivity, Speci.city, Positive predictive value, Negative predictive value,
Nomogram, ROC curve, McNemar chi-square test.
- Software: SAS and R (data management, visualization, analysis); Packages: ggplot, Caret,
descr, exact2x2.
.
Business Analytic: Textual Analysis, System Performance, Key Perfor-
mance Indicator, Share Point Analysis
Tag: Text mining, Wait time analysis, KPI.
Project Title: Cancer Patients Journey Evaluation - CancerCare Manitoba.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Wait time analysis of various steps from diagnosis of disease to receiving
treatment of the disease.
- Application: Visualization and managing data and provide weekly, monthly and yearly reports
for CancerCare Manitoba and Manitoba Health.
- Software: Excel, SAS and R.
.
Drug Utilization and Evaluation
Health Tag: Lung Cancer, Cisplatin, Carboplatin, Bisphosphonate.
Statistics Methodology Tag: Kaplan Meier, Competing Risk, Cox-Proportional
Hazard, Time varying Covariate, Logistic Regression, T-test, Chi-Square Test, Log
Rank Test.
Project Title: The E.ect of Cisplatin Versus Carboplatin on Cancer Outcomes for small Cell Lung
Cancer.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Finding e.ectiveness of two major cancer drug Cisplatin and Carboplatin
on Cancer Outcomes for small Cell Lung Cancer.
- Application: Complete exploratory and survival analysis, overall survival, progression free
survival and multivariate logistic modeling. Preparation of manuscripts for scienti.c publications.
- Software: SAS and R.
Project Title: Bisphosphonate Use and Cancer Risk.
Organization, Duration & Role: CancerCare Manitoba , 2016 - 2017, Biostatistician.
- Project Synopsis: Bisphosphonates (BPs) are a class of drugs known to have anti-tumorigenic
properties. Major goals of the study are to investigate the association between BPs uses and
reducing the risk of cancer incident, and association between BPs uses and survival after cancer
incidents.
- Application: To achieve the goals of the study, time dependent competing risk model (CoxPH
reg) has been used.
- Software: SAS and R (data management, visualization, analysis); Packages: SAS[Proc sql, freq,
sort, univariate, lifetest, phreg, sgplot, gplot], R [ggplot, coxph, surv.t].