VARYASYON KATSAYISININ BAZI ÖZELLİKLERİ VE TÜRKİYE ZİRAİ ARAŞTIRMA KURUMLARINDA YAPILAN DENEMELERE UYGULANIŞI

Abstract

ÖZET
Benzer muamdede ve materyal ile yapltan deneme~
lerden hesaplanan varya~yon katsayıarı belirli bir ortalama
etrafında toplanırlar. Buna göre, Türkiye şartlanda
çeşitli bitkilerle ilgili ortalama varyasyon katsayısı
ve güven sınırları bilindiği taktirde, yapılacak
denemelerin varyasyon katsayıları bu değerlerle karşılaştırılarak
denemenin doğruluk derecesi hakkında bir
hükme varılabilir. Bu hükümde denemenin etkinlik değeri
de yardımcı bir kriter olarak kullanılmalıdır. ,
Bu çalışmada, Zirai Araştırma Kurumlarında uygulanmakta
olan deneme planlarının doğruluk derecelerinin
nasıl ölçülebileceği araştırılmıştır. Ayrıca" denemelerde
hassasiyetin artırılmasına tesir eden faktörler
ilebunların tayininde kullanılan metodlar da incelenmiştir.
Elde edilen sonuçların anlaşılabilmesi ve uygulamayla
ilgilerinin iyi kurulabilmesi için varyasyon kat·
sayısının frekans dağılışı ve güven sınırları gibi teorik
özelliklerine de yer verilmiştir.

SUMMARY

SOME PROPORTlES OF COEFFICIENT OF VARIATION
AND APPLICATINON ON THE EXPERIMENTS CARRIEO OUT
BY TURKISH AGRICULTURAL RESEARCH CENTERS.

The purpose of this study was
to tedermine the comparative sensitivity
of the maıor experimental
designs applied at the Agrieu!tural
Experiment Statiorl$ in 1urkey. Some
factors effecting the accuraey
of the -experiments' and methods
used to identifty these factors
wereconsidered.
As a qui.de for planning the
next experiments confidence interva/,
s of variation coefficients of de·
signs and eertain treatment gro·
ups are obtained· For thE'se proce~
dures the probability distriqution
of the coeffieient of variatin must
be known.
Approximate distr.ibutions have
been deva·loped by McKay (1932)
and Hald (1951) among which the
former's seems tu be more eonvanient. The eXact sampling distribution
ware also obtained by Iglewicz
(1967).
Data used here, ir'ı this study i
were colleeted from various Agri·
cultural Experime:ı-ts as publishedresults
of tre experiments performed
since1960. Contidenee limits
of the coefficionts of variation we·
re estimated and' the best estimates
were obtained uSilıg the nor.mal
distr.ibution, Jakcnife tecnıque
McKay's and Hald's appro·
ximations.
Coeffieients of va! i,ation and
their confidenee intervals found
for different experiments conducted
with different crops and fertİlizers
are shown in the following
table.
---.-..,._--
-------_.._-- v (%) Confidence interva! (% 99)
Wheat
Barley
Com
Gotton
Oat
Wheat fertilizıng
Cotton fertilizing
The eoeffıcıent of var,j'ation of
the experiment conlucted with a
cartain crop palenı must stay, for
the sake of sentivity, under the
upper limit found with the precission
experiments. For example, if
15,04 1231 < V < 19.59
14,72 9.28 < V < 20.16
18,13 13.92 < V < 22,41
9.45 6,61 < V < 13,46
20,98 12,79 < V < 2\3,89
19,03 14,31 < V < 27,77
15,22 11,86 < V < 18,62
---_._----._.
we should and up with an estimation
more than 20 % of the coeffici.
ent of variation we would lake
our chooce of plan'ing of experiment
quite a less sensitive one.
165
in the experiments conducted
with different crops and fertilizers
we tried to compare two methods
of estirnatian of the number of
replicates, namely, the method developed
by Cochr~r;ı and Cox and
that 'by Harris-Horvitz-Mood.
Both of the methods gave the results
with almost equal preciseness,
but conside1ing the eace usability
we would sugge~t to use
the method of Cochran and' Cox
prefarably. Accordingly we found
the numbers of re~licat€s to test
25 % of the treatment means differences
with a level of 20 % type
ii error for the e>:periments with
wheat, barley corn, cctton, aat
crops and wheat and cotton fertilizıng
5, 7, 3, 8, 7, 5, respectively.

Keywords

• -