Presentation Title

Development of MATLAB Codes for Meta-Analysis Applicable to Medical Research Data

Faculty Mentor

Syed Zaidi

Start Date

23-11-2019 8:45 AM

End Date

23-11-2019 9:30 AM

Location

80

Session

poster 2

Type of Presentation

Poster

Subject Area

biological_agricultural_sciences

Abstract

In the field of medical research, multidimensional research efforts lead to data that very often predict conflicting results. For a reader, it becomes hard to make conclusive observations on the outcome of research efforts emerging from various groups. Meta-analysis emerged as a powerful tool for synthesizing complex data coming from various studies. It assists to comprehend the full scope of the study and its impact on the subsequent decisions that are based on final research outcomes. In this work, we identified the need of developing open source MATLAB models that can be used to perform meta-analysis on the given data originating from various resources. For this purpose, three MATLAB models are developed. These MATLAB models perform meta-analysis synthesis on the data by using random effect and fixed effect models. The fixed effect model includes analysis for both continuous and binary data. For this purpose, a corrected standardized mean difference (Hedge’s g) and odd ratios were used to measure effect size for continues and binary data respectively. In the third model, summary effect was computed using random-effects model that in contrast to fixed effect models, allows the true effect to vary from study to study. Three MATLAB models were validated by applying them on various examples on medical data analysis presented by Borenstein et.al [2007]. In each case, size effect, weight assigned to each size, estimation of summary effect, Z value, 95% confidence interval, and P values were calculated. Forests plots for each study were also developed. Results show that three MATLAB models are robust and may be used to implement on a larger scale. The work presented in this conference will include comprehensive details of each model and will present validation results for various studies to ensure the robustness of these models.

This document is currently not available here.

Share

COinS
 
Nov 23rd, 8:45 AM Nov 23rd, 9:30 AM

Development of MATLAB Codes for Meta-Analysis Applicable to Medical Research Data

80

In the field of medical research, multidimensional research efforts lead to data that very often predict conflicting results. For a reader, it becomes hard to make conclusive observations on the outcome of research efforts emerging from various groups. Meta-analysis emerged as a powerful tool for synthesizing complex data coming from various studies. It assists to comprehend the full scope of the study and its impact on the subsequent decisions that are based on final research outcomes. In this work, we identified the need of developing open source MATLAB models that can be used to perform meta-analysis on the given data originating from various resources. For this purpose, three MATLAB models are developed. These MATLAB models perform meta-analysis synthesis on the data by using random effect and fixed effect models. The fixed effect model includes analysis for both continuous and binary data. For this purpose, a corrected standardized mean difference (Hedge’s g) and odd ratios were used to measure effect size for continues and binary data respectively. In the third model, summary effect was computed using random-effects model that in contrast to fixed effect models, allows the true effect to vary from study to study. Three MATLAB models were validated by applying them on various examples on medical data analysis presented by Borenstein et.al [2007]. In each case, size effect, weight assigned to each size, estimation of summary effect, Z value, 95% confidence interval, and P values were calculated. Forests plots for each study were also developed. Results show that three MATLAB models are robust and may be used to implement on a larger scale. The work presented in this conference will include comprehensive details of each model and will present validation results for various studies to ensure the robustness of these models.