Document qmOLOgm1MwpeXzYy9eZJO0QdK
BFGoodrich
The BFGoodrich Company 3925 Embassy Parkway Akron. Ohio 44313
July 27, 1987
Dr. Harold L. Kaplan Southwest Research Institute P.O. Drawer 28510 6220 Culebra Rd. San Antonio, TX 78284 Dear Hal: As you may recall we decided to ask Dr. Sara Debanne to review the statistical analyses that we have used. Her comments are enclosed; please review these so that we can discuss them. Sincerely,
Robert K. Hinderer, Ph.D. Manager, Toxicology jp 0727-1
T
2177001
CASE WESTERN RESERVE UNIVERSITY CLEVELAND, OHIO 44106
A > ' -"7-i / i /
JUL 2 ? 7987
vUL, Q3
/ c, d J
l)ccn
l~li2AX t fl '
o 'hiUo halt
CO ? k c -u ^ cc/ 'U c/<
cW ^lc
H JJWL j^,4^ j j .^ Cc( /sxjj^-ci
k-0 ff <-f
CCv+ j, 6tXc c$
n
do
U1A.
usul
/> ^ fk O
<d
{/U. V/U <l.c lilLlCi,
^ cca.'lcJL^
^ ctvoc g,i4oiA 4 <^-6^
fh cLo'M. j id /o/~u- fV
eo f.~< fU c
c/
42 Lc -t.Oi
fA. d iti d-'
^ .u o'
'-/U
Hux> t/? fliCcLLci 1o
cLci/i^
<yj~ d' Il^jl
J^T ~tmiz ouie/j
V
Oj oV ^cf
/zx^ux'iX
T1 )? "UL-l C7~fc aa( O f-
OU ^(lAJ/nd. /U-j~o
U-f Hjl fleJoi/fl
Q/UJ-*
\XUi T/-< tw-
r u CtacC^
/ilA-w4 j
O "fj-0 o^j> /Li) a cC. oc< k
^ ai k/ k
/OC^ctt/.
c a 4x L J? U '
/
*
T/e
/L i/j tcj -fvo 4
r\ji Vi Ccajic
ot
ci/
J &o'/' ^
^a iOu W' c{
Z'&a.
&-/
A-AxiJc
d te-u
'y
Qu 1
<r- S, ft 'C
CO /
J c 2x^ < ucJLC^
uUCiidd /\l-Oj Ujl S0 j
oi?e^i ^
o /
; )4 [itJL( CJU^* j `Ti ( t
U-cC /d CcV cx/
i
-i
/ 5-c X< r ^ ?\A G. v-\
HmsQ
Ci t'd Jt *fj /I, i ( .
1 luXtA 4. ^ `-T* .
Department of Epidemiology and Biostatistics School of Medicine
2119 Abington Road
y'
<~ c, 6<c
Telephone: (2161 366-3195
ZmLLTZZ
Statistical Review of "Effects of HC1 on Respiratory Response and Pulmonary Function of the Baboon."
The main statistical comments are summarized in detail in the attached TABLE. The several comments of mine in the critique of the TABLE are based on a few issues appearing in various combinations. Here, they will be addressed in a general sense.
1. Transformations of Data
Some dependent variables used in the statistical analyses were derived by mathematical transformation of more original outcome data. The variables, respiratory rate, tidal volume and minute volume, underwent transformation in the analyses of respiratory response (ratio during exposure to baseline), and C0,, challenge response (slope during ramping period and ratio of maximum value during exposure to baseline average).
It could be suggested that the resultant transformed variables have distributions further from the Normal than do the original data and for the respiratory response tests the use of baseline value as a covariate might be more appropriate.
The danger here is that the p-value of a test may be understated (i.e., more likely to result in rejection of a true null hypothesis than presumed) through the use of such transformed data.
2. Statistical Tests
The critique of statistical test utilized involves some subtle modeling points. First, is the general notion that the potential interdependence of outcome variables suggests that the multivariate ANOVA (MANOVA) model structure is more appropriate for analyzing the data than are the separate ANOVA's of the paper. Second, is the comment that 1-way tests (MANOVA's) with repeated measures are preferable to the 2-way ANOVA's presented in much of the paper. Indeed, it is inappropriate to model time as a main effect in a 2-way ANOVA when parallel values over time correspond to values from the same animal (repeated measures).
Along with the TABLE are further ADDITIONAL COMMENTS addressing various iterns.
f:
T
eoozzTz
-a
r- 03
03 03
< 4-> r-
3 > r0 -O
03 +->
o 03 03
to JZ
z a. a
to CD
< 03 03
03 o 4- 4-
OE
03
03 03 SZ 4-
3<
o> >
3 O'
3o Oz
<
+-> +-> CL 03 p~ p' 3 OJ 4- -O 03 <C
+> > 5* +->
03 r* o -c:
4o
E >> 4- o 5 03
+-> Z CD r-- < -r3S E
4- E
to *-H ^
>1 V)
3 03
* 03 03
03 * L
+J 2 &-
4- O 03
o1 3
M to 4--
3 i-H tO
03 03
to 4- a
4- 4- 03
HH m CL i--4 O E
03 -M M to
03 03 03 +> O. 03 03 4-
-Q 03 +-> 4r-- 03 3 4-
03 < 4> CL
o z 03 c s: to
>> 03 to
03 1 4W3
to 03 03 SZ 03 H-- E
"O L-> cu to +-> 03 03 +-> 03 CL 03 03 4. jO
OJ SZ 4-
03 r~ 4-
03 S< Cl > o 03 z < to
>, f0 to
03 41 3 to 03 to x: 03
E
>>
ro to
3 03
l $w <
to --1 zz >
4- 03 to o
O =3 03 03 z
r- _Q 03 <
03 03 E 2: cd
to > *o
c
3 TD X -i-
r- Q. 3 CU 03 M
03 O 4-> t-- 4->
C P (U c ro CL
o ro -4-> to CU E 4-
r-- r- OJ
CL o
4-> 4- t-> r- 03 (J 03
OJ CL tO 03 4.
-Q
E o s-t>
rO
4- 4- 03 O -C
4->
O Cl "U E 4-> * ^
4- CL C
03 CD
(O ro 13 5 tO *r--
cOE
03 --
C0 < 3
4- 03 ro
> 4- f--
4-> Q. E c o 03 03
O 03 z .c -o
to r-- 4-> > < +-> o
co i-- LtJ SI O E
r--
o -Q to r--
03 r--
r-- to
JD
03 to
OJ
o 4-
>,Q_
03
OJ o 4-
E CU 03 4-> 4-
C to 03 CL
o 03 03 3 CL 03 r-- 03 X3
03 03 4-
E>
"O
S- jC r--
o ca +-> 3 4- r-- O
to 03
5
c +->
03 OJ < to
4. M > 03
4- to O 4_
4- z 3
to 03 c to
-a 2: rO
x: c
03
1-- 3 rO E
03
3cr
c
to o>
u 03
to* *cr-
- Q.
<3
to +>
cu E
to
to >1
> O
Oz so->
<C c
E o o
ro
c <
>> rO 03 O
5C I =3
"O r-t Q
c
03
CO
>> c
CO 03
03
CU
E
to "O c
< 03 1
>
o <D 03
Z 4<3C
tO -f-- >0 (0 ro a E
3I XCU to
CSJ 03
to
< > o z <
>i tO 3 i c\j
x: +-> to
r* 03
4-
3
to to
03 < 03
to
>E
c
O0 z -O 03 to <c CU r-- r-
-M CL 4.
>. ro *- rO 03 03 4-> CL
5 CLf-- E
03 3 0
CsJ 4- 21 u
CD 3
x:
to
03
3 4-
3
to to - OJ
< 03
to
>E
c
O0 z "O 03 to
<: 03 r--
+-> CL 4-
>> ro *r- ro OJ 03 +> CL
3 Q_ 1-- E 1 03 3 0
CSJ S- 21 0
<U 43 to
o
03 Q_ 4-> X '--
CO 03 cu
'c
O r--
r- CU
+-> to
SZ o0r3 -0o3
03
>0
OJ 03
E E
to +->
<u rO
3 4. * 03
CO .o 03 E
03 > E 3 4- 3 r--
O r-- O 03 +-> O > > rO >
4- 03
r- r- +->
CL 03 3
tO "O C
03 -r- p*
CC M E
to
ro CD
CO
-O 03 O CD O ro r-- 4SZ 03
>
r0 OJ
03 4- r-- 03 CL +-> E 4- ro < to
to "O 03 O 03 O 03 r-- 4X3 03 to > 03 r-- 03 r-- 03 SZ 03 ro 4- r-- r" 03 CL 4- +-> E 03 4- ro > 03 to
1-- "O to l_i_ c ro Q_ 03 CD
03
4-
a 3 TD
T3 (U O
Oo .T4--
r- CU CO CL
CU
03 4- *03
CU E >> E 3
4O- *--3 rO--
+-> O > 03 >
u <u
Q00. "rOO 3C
CU .r- .rcc *-> E
E 0 .--. 3 +-> 03 E CD r-- 03 ro X 3 4ro r-- 03 E 03 >
> rO 4-
O 03 03 ---- 4- C
3 1-- O to 1-- 1-- O 03 +-> CL to OJ X 03 a: 03 _a
cu to 3 O CL to <u
cu
3 to >0 to
03 4-
CL to 0> er:
T
3 O
O 3 3
>> a 03 3 O to
-- to 3 CU
a. -m
cu CD
3 CU ---- CO
i-- cu 03 tO
.3 3 OO
Cl
CNJtO
o cu O 4-
fO PO
vj
o
Additional Comments 1. On pages 4 & 5 there is discussion of a change in the control of
exposure. Possibly, it would be appropriate to conduct a test of homoscedasticity (e.g., Bartlett's, Cochran's tests) for data generated under each method of control. Heteroscedasticity could lead to understating p-values. 2. On page 14 it appears to indicate that raw response rate data was analyzed by a statistical test, the result of which was non-rejection of the null hypothesis of equality of exposure groups. It is then stated that an ANOVA of slopes during ramping time was conducted whereby some significant differences were detected. Only the second of these statements seems to described on page 10. 3. A textual comment referring to page 10, second paragraph:
LINE 2 of that paragraph replace "with" by "against." LINE 3 of that paragraph replace "between" by "to each other across."
T
2177005