Non-Parametric Test for Ordered Medians: The Jonckheere Terpstra Test
Pages 203-207
Arif Ali, Abdur Rasheed, Afaq Ahmed Siddiqui, Maliha Naseer,Saba Wasim and Waseem Akhtar
DOI: http://dx.doi.org/10.6000/1929-6029.2015.04.02.6
Published: 21 May 2015

Abstract: In clinical trials, sample size is usually lesser as compared to other epidemiological studies to make it more feasible and cost effective. Small sizes of such trials discourage the use of parametric test due to violation of the assumption under which they are applicable. Therefore, the use of nonparametric test is substantial in clinical trials to test two or more independent samples. The Kruskal-Wallis h test is an alternative to one-way ANOVA and can be used to identify significant differences among different populations. When we have several independent samples and assumed to be arranged orderly, Jonckheere Terpstra test is a best choice to compare population medians instead of means. For the application of Jonckheere Terpstra test the data from the study of cleaning methods for ultrasound probes are used. The Jonckheere Terpstra test is recommended over Kruskal-Wallis h test as it compares and provides significant difference between more than two population medians when they arranged in order. Therefore, the aim of this research paper was to explore the use and significance of Jonckheere-Terpstra test with the use of practical example.

Keywords: Jonckheere Terpstra test, non parametric test, comparison of medians.

On the Relationship between the Reliability and Accuracy of Bio-Behavioral Diagnoses: Simple Math to the Rescue
Pages 172-179
Dom Cicchetti
DOI: http://dx.doi.org/10.6000/1929-6029.2015.04.02.2
Published: 21 May 2015

Abstract: An equivalence between the J statistic (Jack Youden, 1950) and the Kappa statistic (K), Cohen (1960), was discovered by Helena Kraemer (1982). J is defined as: [Sensitivity (Se) + Specificity (Sp)] – 1. The author (2011) added the remaining two validity components to the J Index, namely, Predicted Positive Accuracy (PPA) and Predicted Negative Accuracy (PNA). The resulting D Index or D = [(Se + Sp) + (PPA + PNA) – 1] / 2. The purpose of this research is to compare J and D as estimates of K, using both actual and simulated data sets. The actual data consisted of ratings of clinical depression and self-reports of gonorrhea. The simulated data sets represented binary diagnoses when the percentages of Negative and Positive cases were: (Identical; Slightly varying; Mildly varying; Moderately varying; or Markedly varying diagnostic patterns, For both the diagnosis of clinical depression, and the self-reports of gonorrhea, D produced closer approximations to Kappa. For the simulated data, under both identical and slightly different patterns of assigning Negative and Positive binary diagnoses, K, D and J produced identical results. While J produced acceptably close values to K under the condition of Mild discrepancies in the proportions of Negative and Positive cases, D continued to more closely approximate K. While D more closely estimated K under Markedly varying diagnostic patterns, D produced values under this extreme condition that were closer than would have been predicted. The significance of these findings for future research is discussed.

Keywords: Binary Diagnoses, Diagnostic Reliability, Diagnostic Accuracy.

Observation-Driven Model for Zero-Inflated Daily Counts of Emergency Room Visit Data
Pages 220-228
Gary Sneddon, Wasimul Bari and M. Tariqul Hasan
DOI: http://dx.doi.org/10.6000/1929-6029.2013.02.03.7
Published: 31 July 2013

Abstract: Time series data with excessive zeros frequently occur in medical and health studies. To analyze time series count data without excessive zeros, observation-driven Poisson regression models are commonly used in the literature. As handling excessive zeros in count data is not straightforward, observation-driven models are rarely used to analyze time series count data with excessive zeros. In this paper an observation-driven zero-inflated Poisson (ZIP) model for time series count data is proposed. This approach can accommodate an autoregressive serial dependence structure which commonly appears in time series. The estimation of the model parameters by using the quasi-likelihood estimating equation approach is discussed. To estimate the correlation parameters of the dependence structure, a moment approach is used. The proposed methodology is illustrated by applying it to a data set of daily emergency room visits due to bronchitis.

On the Translation of a Treatment's Effect on Disease Progression Into an Effect on Overall Survival
Pages 72-78
Steven M. Snapinn and Qi Jiang
DOI: http://dx.doi.org/10.6000/1929-6029.2015.04.01.8
Published: 27 January 2015

Abstract: There are many examples of treatments for cancer that show a large and statistically significant improvement in progression-free survival (PFS) but fail to show a benefit in overall survival (OS). One recent example that has received considerable attention involves bevacizumab (Avastin) for the treatment of breast cancer. While it seems logical that slowing the rate of progression of a fatal disease would translate into an increase in survival, it is not clear what relative magnitudes of these two effects one should expect. One potential model for the translation of a benefit on disease progression into an OS benefit assumes that patients transition from a low-risk state (pre-progression) into a high-risk state (post-progression), and that the only impact of the treatment is to alter the rate of this transition. In this paper we describe this model and present quantitative results, using an assumption of constant hazards both pre-progression and post-progression. We find that an effect on progression translates into an effect on survival of a smaller magnitude, and that two key factors influence that relationship: the magnitude of the difference between the hazard rate for death in the pre- and post-progression states, and the duration of follow-up.