The c ho ices of experim enta l des ign as well as o f statistical ana lys is are of great importance inrnfi e ld experiments. To improve production and productivity of crop yield it is important tornemploy tested and proven exper imental des ign. The aim of th e study is to eva luate the prec is ionrnof commonly used experimental des igns (randomi zed complete block des ign, lattice des ign andrnalpha- lattice des ign) in the Ethiopian agricultural research system. Both fixed and linear mixedrneffect mode ls were used to compare the effi ciency of experimental des igns. Different varietyrntrials co nducted at different locations by the National and Regio nal Agricultural ResearchrnOrganization in 2004/05-2008/09 years (seasons) were used in this study. The results indicaternthat on average the randomized complete block design (RCBD) was more effic ient thanrncomplete ly randomized des ign (CRD). For large miss ing va lues in RCBD, the use incompleternblock approach is more e ffi c ie nt than using miss ing-estimated approach. A Iso use 0 I' covariancernana lys is for damaged yield in RCBD increased the re lative effic iency of RCBD compared tornCRD than the ana lys is without cova riate. The lat tice des ign was more effic ient to increase thernprec ision o f agr icultural fie ld experiments than RCBD. The re lative e ffi c iency o f a lpha latt icerndes ign compared to RCBD was 101.9% and 102.4% for both data sets respect ively. Th isrnindicates that researchers ha ve in gcneral correctly co ntro lled the loca l va riability duringrnexperimentat ion. However, the cases where blocking was not s ignifica nt can be attr ibuted torntechnical issues such as orientation a nd di rection of blocking which should be improved byrnresearches in similar experiments. Based on the results we co nc lude that RCBD is more efficientrnthan CRD and lattice and a lpha lattice des igns are more effic ie nt than RCBD. In order tornincrease the precision of agricultural field experiments researchers are advised to use RCBD forrnsmall number of treatments, lattice des ign and alpha lattice desig ns for large number ofrntreatments.