Home
Tags
No Tags
ASSESSMENT OF AGREEMENT AND SELECTION OF THE BEST INSTRUMENT IN METHOD COMPARISON STUDIES
| Title: | ASSESSMENT OF AGREEMENT AND SELECTION OF THE BEST INSTRUMENT IN METHOD COMPARISON STUDIES |
| Author: | Choudhary, Pankaj K |
| Description: | We consider three problems that arise in studies comparing two or more methods that measure a continuous variable of interest. First is the basic problem of how to assess the degree of agreement between two methods (or instruments). If there is sufficient agreement between them, one can use them interchangeably or prefer the one that is cheaper or is easier to use. We employ tests of hypotheses to assess if the data have evidence for satisfactory agreement. In particular, we consider three formulations of satisfactory agreement and provide tests for the associated hypotheses. The formulations are: (i) both the mean and the standard deviation of the difference of measurements from the two instruments are close to zero; (ii) the means of the two measurements are close, their standard deviations are close, and their correlation is high; and (iii) a large proportion of the difference lies in an interval close to zero. In all the cases, the thresholds determining closeness are supplied by the investigator. The first two formulations give additional information regarding the nature and extent of disagreement when the data do not have evidence for satisfactory agreement. For the last formulation, we discuss both the parametric and nonparametric tests. Comparison of two instruments with a gold standard is the focus of the other two problems. We first consider how to select the instrument that agrees most with a gold standard in terms of mean squared deviation. This instrument is designated as the best one. For this selection problem, we present two large sample single-stage procedures and a two-stage procedure using the multiple comparisons with the best approach. Questions like which parameterization works well for the comparison and what sample sizes are adequate are answered using asymptotic theory and simulation. For the last problem, we determine, through a hypotheses test, whether the best instrument agrees sufficiently well with the gold standard before proceeding to its selection. We describe a two-stage procedure for this purpose and study its properties using the asymptotic distribution of the test statistic and simulation. |
| Permanent Link: |
http://rave.ohiolink.edu/etdc/view?acc_num=osu1029109764
http://hdl.handle.net/2374.OX/7986 |
| Date: | 2002 |
Files in this item
| Files | Size | Format | View |
|---|---|---|---|
|
There are no files associated with this item. |
|||