Categorical data analysis
Presenter: Christopher Bilder
When: Friday, July 15, 2016 Time: 9:00 AM - 10:30 AM
Room: Saanich 1-2 (Level 1)
Session Synopsis:
Objective functions for group testing
Specimen screening for infectious diseases often uses group testing in high volume settings. Rather than testing specimens one by one, group testing works by pooling specimens, such as blood or urine, from separate individuals to form a single specimen. Individuals within negative testing groups are declared negative. Individuals within positive testing groups are retested in some predetermined manner to distinguish the positive individuals from the negative ones. As long as disease prevalence is small, group testing will greatly reduce testing loads when compared to individual testing. To achieve the maximum reduction in these loads, the choice of group sizes is extremely important. These sizes are typically chosen based on optimizing an objective function that involves the expected number of tests, assay characteristics, and/or laboratory constraints. The purpose of this presentation is to evaluate different objective functions and compare their use among a number of group testing algorithms. For situations that most often occur in practice, we show that actual group testing implementation would be largely the same for currently used and recently proposed objective functions.
Categorical data analysis
Presenter: Youyi Fong
When: Friday, July 15, 2016 Time: 9:00 AM - 10:30 AM
Room: Saanich 1-2 (Level 1)
Session Synopsis:
Nonlinear Models for Immunological Assay Outcome from Two Dilutions
Serial dilution assays with continuous experimental outcomes are often used to quantify substance in a biomedical sample. The design and analysis of serial dilution assay data has so far primarily been based on estimating dilution-response curves, the nonlinear relationship between sample dilutions and experimental outcomes. Motivated by the need to improve the measurement of antigen-specific binding antibody concentrations in HIV-1 vaccine studies, we study the nonlinear functional relationship between experimental outcomes measured at two different dilutions, which we call paired response curve (prc). Zhang et al. (2009) first proposed a three-parameter model to study this type of functional relationship in the context of reverse phase protein array (RPPA) in tumor genomics. In this paper, we propose a more general, four-parameter prc model. Estimation of paired response curve involves a diverging number of incidental variables, a fact that was ignored by Zhang et al. To properly handle errors-in-variables, we propose a total least squares estimator, which profile out the incidental variables and is equivalent to the maximum likelihood estimator of the functional relationship. We study the asymptotic theory of the estimator to derive an analytical variance estimate. We then show how to use paired response curve to predict experimental outcomes at novel dilutions. Our simulation studies show that the proposed four-parameter model and total least squares estimator have good finite sample performance in both estimation and prediction. We apply prc to a real data example from a study of maternal immune responses and Mother-To-Child-Transmission of HIV-1 (MTCT) (Permar et al. 2015), and show that prc fits the data well and performs much better at prediction than linear intrapolation.
Categorical data analysis
Presenter: Graham Hepworth
When: Friday, July 15, 2016 Time: 9:00 AM - 10:30 AM
Room: Saanich 1-2 (Level 1)
Session Synopsis:
Estimation of proportions by group testing: retesting revisited
In group testing, units are pooled together and tested as a group for the presence of an attribute, such as a disease. If the test result is positive, it is assumed that at least one unit in the group is positive, and if the result is negative, that all units are negative. Cost savings can be considerable if the population prevalence is small. Group testing for estimation of proportions has been applied in a wide range of contexts, including estimation of virus prevalence in flowers, presence of genetically modified plants, fish and wildlife disease prevalence, and transmission of viruses by mosquitoes. When estimating proportions (as opposed to identification of positive individuals), retesting of units within positive groups has not received much attention. This is largely due to the minimal gain in precision when compared to testing additional units. However, If acquiring additional units is impractical or expensive, and testing is not destructive, we show that retesting can be a useful option. We propose the retesting of a random grouping of units from positive groups, and compare it with nested halving procedures. Using simulation, we show that for most realistic scenarios, our method is more efficient than other retesting methods. Hepworth G & Watson RK. (2015) Revisiting retesting in the estimation of proportions by group testing. Communications in Statistics – Simulation and Computation DOI: 10.1080/03610918.2014.960725.
Categorical data analysis
Presenter: Claudia Rivera
When: Friday, July 15, 2016 Time: 9:00 AM - 10:30 AM
Room: Saanich 1-2 (Level 1)
Session Synopsis:
On the analysis of two phase designs in cluster-correlated data settings
In resource-limited settings researchers often only have access to aggregated or group-level data. Analyses based on such data are well-known to be susceptible to a range of biases, collectively termed ecological bias. Unless one is willing to make untestable assumptions, the only reliable approach to valid estimation and inference is to collect a sub-sample of individual-level data. One cost-efficient strategy is the two-phase design in which the population is initially stratified by a binary outcome and categorical variable or combination of categorical variables. While methods for two-phase designs are well-known, to our knowledge they have focused exclusively on settings in which individual study units are independent; that is, no methods exist for the design and analysis of two-phase designs in cluster-correlated data settings. To fill this gap we develop for valid estimation and inference, the first based on inverse-probability weighting (IPW) and the second on a pseudo-likelihood. For both methods, user-specified working correlation matrixes can be specified with inference based on a robust sandwich estimator. For the IPW estimator we develop an calibration algorithm that makes use of the readily-available group-level data to improve efficiency. A comprehensive simulation study is conducted to evaluate small-sample operating characteristics of the proposed methods. Finally, the methods are applied to a large implementation science project examining the effect of an enhanced community health worker program to improve adherence with WHO guidelines for at least 4 antenatal visits, among nearly 200,000 pregnancies in Dar es Salaam, Tanzania.
Categorical data analysis
Presenter: Hideaki Uehara
When: Friday, July 15, 2016 Time: 9:00 AM - 10:30 AM
Room: Saanich 1-2 (Level 1)
Session Synopsis:
Intrasubject parallelism in the paralle-line bioassay
In this presentation we compare the recently proposed criteria to evaluate the intrasubject parallelism, which is the essential assumption to evaluate the relative potency using a parallel-line bioassay. The first metrics we consider is the aggregated criterion which was derived via the translation of individual bioequivalence criterion stated in the regulatory guidance (Food and Drug Administration, 2001). The second is a single criterion based on the intrasubject slope ratio which uses the approximate two one-sided tolerance limits of linearized variables. These procedures are demonstrated in an example analysis, and their properties are evaluated through Monte Carlo simulations. The power of aggregated criterion was shown to be high for designs of moderate size. The tolerance limit approach could give more interpretable results, in spite of its lower power. Extensions to the parallel-curve assay and the integration of multiple assays will also be discussed. [Reference] Uehara H., Satoh K., Komokata N., Tokita Y., Nishiyama M., Iida M., Ishikawa J., Ogawa K. and Yamamoto M. (2016). Combinability of animal data in relative potency estimations. Submitted.
