jerry
☆    

Russia,
2015-08-26 10:37
(3137 d 14:57 ago)

Posting: # 15332
Views: 13,179
 

 can't get information about outliers detection [🇷 for BE/BA]

Hello) It's me again. When i performed 2x2x4 design i didn't get analysis of outliers detection
I decided just use your example from here
http://pkpd.kmu.edu.tw/bear/
testdata.csv with 2x2 design
but after analysis i didn't get output outliers detection
What did you do to get Analysis of outliers detection and related plots?

i did so.
1.load data
2.nca
3.statistical analysis
i have
3537281_2015-08-26_stat_sum.txt
6022843_2015-08-26_nca_pk.txt
3537281_2015-08-26_anova.txt
here this
https://www.sendspace.com/file/9dy3he


Edit: category changed [Ohlbe]
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2015-08-26 10:53
(3137 d 14:41 ago)

@ jerry
Posting: # 15333
Views: 11,909
 

 ODA in bear

Hi jerry

Have you enabled ODA (outlier detection analysis) in your setting? See the picture (may be a little different from yours right now. I am still working on it.). Default is '0' which means ODA 'disable'. Please use ODA carefully. Search the Forum for ODA discussion. BTW, please select the correct category for your post next time. This should be category of R for BE/BA. Thanks.

[image]

All the best,
-- Yung-jin Lee
bear v2.9.1:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
jerry
☆    

Russia,
2015-08-26 18:00
(3137 d 07:34 ago)

@ yjlee168
Posting: # 15341
Views: 11,934
 

 ODA in bear

Dear yjlee168. Yes you are right, if oda=1, then for design 2x2 i got oda stat:)
But if oda=1 and design 2x2x4 i didn't get oda stat.txt))
does replicative design have any limits in Bear for ODA?
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2015-08-26 19:40
(3137 d 05:53 ago)

@ jerry
Posting: # 15343
Views: 11,750
 

 ODA in bear

Dear jerry,

No, ODA is only implemented only for 2x2x2 study with bear. I don't know if there is ODA for replicate design at the moment.

All the best,
-- Yung-jin Lee
bear v2.9.1:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
jerry
☆    

Russia,
2015-08-26 19:58
(3137 d 05:36 ago)

@ yjlee168
Posting: # 15344
Views: 11,877
 

 ODA in bear

That explains a lot
thank you;-)
d_labes
★★★

Berlin, Germany,
2015-08-27 10:12
(3136 d 15:22 ago)

@ yjlee168
Posting: # 15347
Views: 11,693
 

 ODA for replicate designs

Dear Yung-jin!

❝ No, ODA is only implemented only for 2x2x2 study with bear. I don't know if there is ODA for replicate design at the moment.


Have a look into:

Robert Schall, Laszlo Endrenyi, Arne Ring
"Residuals and Outliers in Replicate Design Crossover Studies"
Journal of Biopharmaceutical Statistics, 20: 4, 835-849 (2010)
DOI: 10.1080/10543401003618876

Regards,

Detlew
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2015-08-27 10:51
(3136 d 14:42 ago)

@ d_labes
Posting: # 15350
Views: 11,642
 

 ODA for replicate designs

Dear Detlew,

Many thanks for your info update. Hmm.. '...Removal of single-data-point outliers from analysis may be justified in certain cases.' I should take a look on it.

All the best,
-- Yung-jin Lee
bear v2.9.1:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
d_labes
★★★

Berlin, Germany,
2015-08-27 11:17
(3136 d 14:16 ago)

@ yjlee168
Posting: # 15352
Views: 11,630
 

 ODA for replicate designs

Dear Yung-jin,

❝ Many thanks for your info update. Hmm.. '...Removal of single-data-point outliers from analysis may be justified in certain cases.' I should take a look on it.


You have mail.

But don't invest too much.
I myselve usually don't deal with oulier detection at because by most regulations it's not allowed to exclude data on statistical reasons alone.
At the time of statistical analysis you may not have the possibilities to obtain other reasons which support your outlier finding. Usually it's rather too late for that.
And you usually can't convince regulators that other reasons are not far-fetched.

Regards,

Detlew
yjlee168
★★★
avatar
Homepage
Kaohsiung, Taiwan,
2015-08-27 11:30
(3136 d 14:03 ago)

@ d_labes
Posting: # 15354
Views: 11,804
 

 ODA for replicate designs

Dear Detlew,

❝ You have mail.


Got it. Thank you so much.

❝ But don't invest too much...


Very agree with your points.

❝ ... And you usually can't convince regulators that other reasons are not far-fetched.


All the best,
-- Yung-jin Lee
bear v2.9.1:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear
Download link (updated) -> here
mittyri
★★  

Russia,
2015-08-27 12:20
(3136 d 13:14 ago)

@ yjlee168
Posting: # 15355
Views: 11,713
 

 ODA for replicate designs - "EMA crippled model"?

Dear Yung-jin, Dear Detlew,

You're absolutely right, most regulations are'nt allowed to exclude data on statistical reasons alone.

I explored forum some time ago searching the solution for "EMA Crippled model" (possible solutions for outliers detection during CVwr estimation according to the EMA GL and Q&A).
Very lengthy posts are here, here and here
BTW I didn't find the exact solution for R. Could you please enlighten me?

Yung-jin, it would be great to implement CVwr estimation in Bear according to the European requirements for replicate designs.

Kind regards,
Mittyri
d_labes
★★★

Berlin, Germany,
2015-08-27 12:53
(3136 d 12:40 ago)

@ mittyri
Posting: # 15356
Views: 11,741
 

 ODA for replicate designs - "EMA crippled model"?

Dear mittyri,

❝ BTW I didn't find the exact solution for R. Could you please enlighten me?


Which "exact solution" you are seeking for :confused:.

Regards,

Detlew
mittyri
★★  

Russia,
2015-08-27 13:00
(3136 d 12:34 ago)

@ d_labes
Posting: # 15357
Views: 11,695
 

 ODA for replicate designs - "EMA crippled model"?

Dear Detlew,

I mean the R code for CVwr estimation and outliers detection in Reference dataset according to this point of EMA GL:

"The applicant should justify that the calculated intra-subject variability is a reliable estimate and that it is not the result of outliers."

Kind regards,
Mittyri
d_labes
★★★

Berlin, Germany,
2015-08-27 15:17
(3136 d 10:16 ago)

@ mittyri
Posting: # 15358
Views: 11,630
 

 "EMA crippled model" - CVwR estimation and ODA in R

Dear Mittyri,

❝ I mean the R code for CVwr estimation and outliers detection in Reference dataset according to ... EMA GL:


since I have to do my daily work with SAS I don't have an R solution for doing this out of the box.

But I think it should not that complicated to do the estimation of CVwR and this could be programmed even by a novice in R.
Part I: Estimation of CVwR
  • Make a subset of your data.frame containing only the PK metrics under R(eference)
    dtaR <- subset(data, tmt=="R")
    dtaR$period <- as.factor(dtaR$period)
    dtaR$subject<- as.factor(dtaR$subject)
    dtaR$sequence<- as.factor(dtaR$sequence)

  • Fit an model containing period sequence and subject within sequence
    muddleR <- lm(log(PKmetric) ~ period + sequence + subject%in%sequence, data=dtaR)
    or according to our Captn's preferred implementation simply
    muddleR <- lm(log(PKmetric) ~ period + sequence + subject, data=dtaR)
  • Extract the error mean square from that model and you calculate the CV from it
    msewR <- sum(muddleR$residuals^2)/muddleR$df.residual
    CVwR  <- sqrt(exp(msewR)-1)*100 # in %

Part II - ODA
It is another story and is a grey zone. An outlier test is not expected. So simply create a boxplot from the residuals, cross your fingers and hope that not suspected "outliers" are visible:

boxplot(muddleR$residuals)


This is of course only a rough description just in this moment came to my mind. Intended to give you a starting point. Needs to be tested and validated before going in production. Needs also to make a nice printout of the results. It's up to you.

Regards,

Detlew
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
81 visitors (0 registered, 81 guests [including 10 identified bots]).
Forum time: 00:34 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5