WHEN WILL WE BEGIN TO REDUCE ALPHA AND BETA ERRORS IN STATISTICAL PSI EXPERIMENTS

• Document Type: CREST
• Collection: STARGATE
• Document Number (FOIA) /ESDN (CREST): CIA-RDP96-00792R000700660002-7
• Release Decision: RIFPUB
• Original Classification: U
• Document Page Count: 3
• Document Creation Date: November 4, 2016
• Document Release Date: October 27, 1998
• Sequence Number: 2
• Content Type: OPEN

Approved For Release 2000/08/11: CIA-RDP96-00792R000700660002-7
Papers
star "makes or breaks" the results. By using large "anonymous"
source groups the incentive for any one individual to create false
anomalies might be greatly reduced.
WHEN WILL WE BEGIN TO REDUCE ALPHA AND BETA ERRORS
IN STATISTICAL PSI EXPERIMENTS?
Ulrich Timm (Institut fur Grenzgebiete. der Psychologie and
Psychohygiene, Eichhalde 12, 7800 Freiburg i.Br. , West
Germany)
In many psi experiments some statistical selection errors are
made, after whose correction the initial statistical significance dis-
appears. These are Type I errors, more simply called alpha errors.
That does not necessarily mean, however, that in these experiments
real psi effects do not exist, since the usual methods, if utilized
correctly, are often so ineffective--with regard to the rareness,
instability, and inconsistency of psi effects--that they can only
seldom lead to statistical significance. This inefficiency of statisti-
cal methods creates Type II errors, or beta errors. Therefore,
our objective should not only be the reduction of alpha errors and
the related decrease of spurious significances but also the reduction
of beta errors and the related increase of real significance.
First I give an overview of those alpha errors that I call
statistical selection errors. These show, simply stated, the follow-
ing three qualities (Timm, ZP, 1983, 195-229):
(1) From, a set of statistical results a single result is se-
lected and evaluated by some significance test.
(2) The selection is not performed randomly but according
to a criterion that is related to the level of the single
result in that it directly or indirectly favors positive
results.
(3) Despite this success-dependent selection, the significance
test is carried out and interpreted in the usual manner
without any correction.
Approved For Release 2000/08/11: CIA-RDP96-00792R000700660002-7
 Approved For Release 2000/08/11: CIA-RDP96-00792R000700660002-7
o O 0 u) m m m m w w m m w o
O. 2 O rt ~? m ~~.. m 9 m C-1
a m (D o (D `w o m m
v a~i
m O a DC CD H IC ? an A? :2 y m m
CD 0 J ~y CD CD F. CD o a ro a m B m m F,4 0 6 G? a
W (D
0
0, r- ?
o cn a a w w n 0 CD o z
o ~i
c) CD
ro w a
D
DC ~.
CD o
0
rt N "MD. o
m m ~
m
(D
c_v R
a m p m m m '+ w w k g
0 V, w a w m C m CD
0 o
w
w
o k B o
m w~ C m ?+" m
(D C r,
? m
M, 10 p n rt DIq r< m ro m a ~, a
rr~~
(D m 0 CD O w H n p M phi m m m N (D CD
N
m m o m G ro n m '-y e- a
0 CD (D
(D H
rtm.CD m
0
U w o a n b 0 m w ?n B
m t~ m H FO 9 m (D 4 4 ~. rt
M CD
o CD ~? N m G m m aG m p+
It o R w m
0 3 H O a mro s CIE '0 M m m O
~. O
m rt m m m wan [f 0 m
Uq (D
B' n 0
m m
n - m '
v N
CD pJ
9 C 0
0 CD
a) C
CO
v CO CD
0 (D B'
m m
m
m an ~t m o
O 0
By DC m
,
O (D iH m
0 ca
0 m
a rt m
p ca m 9 I w 0 o
P m m m
q * m w 5C n
o ro w m o o M y m ?m m
ti ro W~j . W o m 'C O 0 B con m m m
m r m * ro a su m m m
m o~q (D 0 0 C rt` w m a1 m m O CD
rt o W w ?.o'-' a 0 o m
H ~. D m m DC
m O o
aq + a m m s o m
m w w a m w m rQ* _ C p C m rt ,_? m h
CO
Oro m 0 CD ~? N N O O m w
t3 0? B m m o (9 O m y aq m 0 O
con u) p ~*?R.m '- 0 m w g - m m - W w ? w 0 w w w (D I
c0 o mC B p co CD D,'d a C w m rt `d O O o 0, O O ID O
0 o mro m. m g CD m m C FS 0 a (D ~r CC m?? o? m p ' m " m -