02/10/2010 — Improvement in Performance by Combining Biomarkers in Diagnostic Medicine

Aasthaa Bansal

When an existing marker does not have sufficient diagnostic accuracy on its own, new markers are sought with the goal of yielding a combination with better performance. Understanding the properties that a new marker should have in order to improve performance would help biomarker development. Assuming the joint distribution of baseline and new markers is bivariate normal in cases and controls, we quantify the improvement in the ROC curve as a function of the correlations and the relative performance of the new marker. The ROC is typically improved substantially only when the new marker performs well on its own and is weakly correlated with the existing marker. Surprisingly, in some realistic settings a very highly correlated marker can yield substantial improvement. We also show that combining markers incorrectly can lead to decrements in performance.

02/03/2010 — Grant Proposals 101

Carolyn Rutter

Grant writing is often a large part of a biostatistician’s career. There are several advantages to controlling your own projects as Principal Investigator, and the demonstrated ability to successfully compete for grant funding is increasingly considered as criteria for promotion in academic settings. In this seminar I’ll review some strategies for such successful competition, focusing on the proposals to the NIH. Topics will include: types of funding mechanisms, submission cycles, proposal format, grant writing style, budget considerations, and the review process.

Grant writing is often a large part of a biostatistician's career. There are several advantages to controlling your own projects as Principal Investigator, and the demonstrated ability to successfully compete for grant funding is increasingly considered as criteria for promotion in academic settings.  In this seminar I'll review some strategies for such successful competition, focusing on the proposals to the NIH. Topics will include: types of funding mechanisms, submission cycles, proposal format, grant writing style, budget considerations, and the review process.Grant writing is often a large part of a biostatistician's career. There are several advantages to controlling your own projects as Principal Investigator, and the demonstrated ability to successfully compete for grant funding is increasingly considered as criteria for promotion in academic settings.  In this seminar I'll review some strategies for such successful competition, focusing on the proposals to the NIH. Topics will include: types of funding mechanisms, submission cycles, proposal format, grant writing style, budget considerations, and the review process.

01/27/2010 – Risk Factors for End Stage Renal Disease in Wilms Tumor Patients

Jane Lange

Wilms tumor (WT) is a pediatric cancer of the kidney that is the sixth most common cancer in children. Children with Wilms tumor are at risk of developing end stage renal disease, due to surgical procedures that remove most or all kidney tissue in those with bilateral WT and because of kidney pathology due to WT treatment (radiation neprhitis, chemotherapy toxicity) or hereditary conditions. Specific groups associated with germline mutations in the WT1 region of the genome (e.g. Denys-Drash and WAGR syndrome), have particularly high risk of developing ESRD. Along with its role as a tumor suppressor gene, WT1 expression maintains kidney function; reduced WT1 expression leads to nephrotic syndrome, glomerulosclerosis, and ultimately renal failure.

Less well described are factors associated with increased rates of ESRD in the patient population without known constitutional WT1 mutations. Using 9,022 patients who participated in the National Wilms Tumor Studies I-V, I present here initial analyses investigating risk factors for ESRD development in patients not in previously described anomaly groups. These factors include bilateral WT status, age at diagnosis of WT, histologic type of the tumor, and the presence and location of precursor lesions consisting of remnants of undifferentiated embryonic tissue. The goal of such an analysis is to provide risk profiles of ESRD in WT patients that may help direct cancer treatment, particularly for patients with bilateral disease.

01/20/2010 — CANCELED

This week’s student seminar has been canceled. Next week will be held as scheduled.

01/13/2010 — On and Off Semiparametric Models

Takumi Saegusa

Finding the efficient estimator has been and will be the center of statistical research. Efficiency, by definition, heavily depends on the statistical model to consider. On the other hand, statisticians seem to love to cite a Box’s famous remark, “Essentially, all models are wrong, but some are useful.” Does the efficient estimator always useful? Is it so under slight model misspecification? We consider the efficient and inefficient estimators that are regular in a semiparametric model and compare asymptotic mean squared errors and asymptotic powers in the local neighborhood of the true model. We geometrically characterize the situation that the inefficient estimator performs better than the efficient estimator with influence functions and “perturbation”, analogue of the score function under local model misspecification, as geometric objects in the space $L_20(P_0)$ of mean-zero square integrable functions with respect to the probability measure $P_0$. In this talk, we use two-phase sampling as a motivative example, and discuss implications of our result to applications.