Agricultural Research Management Information System

Home
Research Summary
Search
About Us
- ARMIS
- Brochure
Contact Us
Report
User Request
Data Input
Help
- Operation Manual
  - PDF
  - Video
- Program Area & Commodity
We have reached 37600 number of research entries at this moment.

- Logout

Research Detail

Home
Research
Detail

Coefficient of Variation as a Criterion of for Using Dmrt

Dr. Debi Narayan Rudra Paul
CSO and Head
Agricultural Statistics Division, BRRI, Gazipur

Abstract/ Executive Summary:

To examine the effect of CV on DMRT and develop a criterion for performing a reliable DMRT if the test is found to be affected by CV. For this study, hypothetical data were generated through computer programmes, using SAS. DMRT can safely be used from comparing treatment means for CV up to 15%. Interpretation of the results of DMRT should be carefully done when sample CV ranges 15-20%. Use of DMRT should be restricted when the sample CV exceeds 20%.

Keywords: CV, Criterion, DMRT,Computer programmes

Location: Agricultural Statistics Division, BRRI, Gazipur

Start Date: 01-07-1989

End Date: 30-06-1990

Program Area: Socio-economic and Policy

Non-commodity: Analysis and interpretations

Objectives:

To examine the effect of CV on DMRT
To develop a criterion for performing a reliable DMRT if the test is found to be affected by CV.

Methodology:

For this study, hypothetical data were generated through computer programmes, using SAS satisfying following conditions- • Designed used was completely randomized design (CRD) with 5 treatments each replicated four times. • Data were generated by using fixed effect linear model Yij=m+tj+ejj. i= 1...5, j= 1...4 was used were m, tj, Yjj ejj stand for general mean effect with treatment effect, observation (ie, yield) in the (i,j)th cell and random errors in (i,j)th cell, respectively. eij s are assumed to be normally and independently dis- tributed with 0 mean and constant variance (cr²) • The true effect oftreatrnents were taken as 1.5(=tl), 2.0(=t2), 2.5(=t3),3.0(=t4), 3.5(=t5) and the general mean effect as 2.5 (=m).• Ten values of o, namely 0.1, O. 2, 0.3, 0.4, 0.5, 0.6, 0.7,0.8, 0.9 and 1.0, were considered and for each er 6 sets of data were generated. In set 1, 1000 experimental data were generated. For each experiment, a random samples of size 20 was drawn from normal population with 0 mean and for a specified value of s. The sample values thus obtained constituted 20 experimental errors corresponding to 20 experimental units. The yield in each experimental unit was then computed following the linear model stated above. The number of experiments in sets 2, 3, 4, 5 and 6 were 10000, 20000, 30000, 40000 and 50000, respectively. The hypothetical experimental data were evaluated by analysis of variance and the treatments were compared by using DMRT. A difference wasscored as 'I' if it followed the proper order of effects, '0' if there was no difference or it was in the opposite order of the true effects. The total of the scores for an experiment was used as the index of reliability of DMRT for the comparison of treatment means.

Source of Information: BRRI Annual Report 1996-97, 170-171, BRRI, Gazipur.

Article Link (Online Journal):

Funding Source:

Total Budget (Tk.) :

Achievement/Findings:

Results clearly certify that sample CV may be used as an indicator of the reliability of DMRT. Provided the assumptions for analysis of variance are tenable for experimental data, DMRT can safely be used with full confidence for comparing treatment means if the sample CV remains within 10%. The level of confidence remains within 95% if the sample CV does not exceed 15%. The results also indicated that one may use DMRT when the sample CV exceeds 15% but should be careful in interpreting results. Use of DMRT, as the results suggest, should be restricted within sample CV of 20%.

Knowledge Management: Report/Proceedings