Simpson paradox

Dear all
I'm an agronomists and i work in a test facility in Italy
In many field trials with insecticides sometimes happen something that seems
to Simpson's paradox
I will try to explain better with an example

One of most important pests od potatoes is Colorado beetle: it born from an
egg and after 4 moulting we have the adults (like a big ladybird but with
row).
For simplicity consider only larvae distinguish within thin (TL) and big (BL).
When i apply 2 different insecticides (A e B) obtain following results of
motlaity:

Insecticides
A B
TL 48 death on 55=87% 100 death on 130=77%
BL 46 death on 88=52% 5 death on 12=42%

So it is allowed say that A insectidice is more efficacious than B either on
thin larvae (87% vs. 77%) than biggest ones (52% vs. 42%).


Now if i sum for colomns....

Insecticides
A B
TL+BL 94 death on 143=66% 105 death on 142=74%

that is just the opposite!!!!!

Some considerations
65% (185/285) of all insects are thin larvae and is well-known that toxycity
is strong influenced by insect age and body dimensions....but wich is the BEST
insecticide A or B???

Wich is the CORRECT statistical model that must be applied???
Where i can find any ciatation of articles that can be founded also in italy
about this argument????

tnx a lot in advance


-------------------------------------------------------------------------------------------------------------------------
Landini dr. Massimiliano
Tel. mob. (+39) 347 140 11 94
Tel./Fax. (+39) 051 762 196
e-mail: robfleming (at) tiscali (dot) it
-------------------------------------------------------------------------------------------------------------------------
Legge di Grelb sugli errori: In una qualsiasi serie di calcoli, gli errori avranno la tendenza a prodursi dalla parte opposta a quella da cui si comincia a controllare se ci sono errori.
-------------------------------------------------------------------------------------------------------------------------
.

Loading