rasheed
☆    

2010-04-14 09:30
(5097 d 07:10 ago)

Posting: # 5115
Views: 11,007
 

 Outlier detection in BE [Outliers]

Dear all
Detection of outlier is very important in bioequivalence study various of statistical methods have been developed to detect outliers in bioequivalence study for 2X2 crossover design, I am intersted in higher order crossover design where number of sequence or periods are greater than treatments to be compared, I am seeking for statistical methods to detect outliers, is there anyone who has some knowledge about my query?

regards
Rasheed.
d_labes
★★★

Berlin, Germany,
2010-04-14 10:29
(5097 d 06:11 ago)

@ rasheed
Posting: # 5116
Views: 9,951
 

 Outlier detection in BE in higher-order x-over

Dear Rasheed,

To :google: or not to :google: that is the question.
Hamlet by William Shakespeare. Don't mix it up with "To Bee or Not to Bee" :cool:

Have a look at

V. Luzar-Stiffler, C. Stiffler
"Higher-order crossover design outlier detection"
27th International Conference on Information Technology Interfaces, 2005.
p 650 - 655
Online resource

and

R. Schall, A. Ring, and L. Endrenyi
"Residuals and outliers in replicate design crossover studies"
Technical report, University of the Free State
Online resource

Regards,

Detlew
Dr_Dan
★★  

Germany,
2010-04-14 12:29
(5097 d 04:10 ago)

@ rasheed
Posting: # 5120
Views: 9,593
 

 Outlier detection in BE

Dear Rasheed
Please excuse my curiosity, but what is the advantage of outlier detection? You just know that you have an outlier but at that stage you are not allowed to exclude this subject from statistical analysis...
Kind regards
Dan

Kind regards and have a nice day
Dr_Dan
rasheed
☆    

2010-04-14 14:57
(5097 d 01:43 ago)

@ Dr_Dan
Posting: # 5125
Views: 9,665
 

 Outlier detection in BE

Dear Dan
yes I know very well that we are not allowed to exlude outliers which are detected on the basis of statistical evaluation, but if this is case so why outlier is the part of FDA statistical guidlines? and why so many researcher has developed different statisitical methods to identify the outliers in the bioequivalence? can you justify it?

regards
rasheed
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2010-04-14 21:11
(5096 d 19:28 ago)

@ rasheed
Posting: # 5138
Views: 9,624
 

 Outlier detection in BE

Dear Rasheed & Dan!

I know that especially European regulators are close to anaphylaxis when it comes to outliers. But whether they like it or not, from a scientific POV:
  1. One has to know if there are any (i.e., not look at the data driven by some gut-feeling – but perform statistics).
  2. Try to find explanations (mistakes in data transfer/coding, randomisation errors, bioanalytical problems,…)
  3. Have a statistical contingency plan in the protocol.
The ‘method’ of taking all data as they are in any case - and screaming ‘Bias!’ if otherwise, is an example of Bad Science. Sorry.

BTW, see the section about outliers in the recent Canadian Draft Guideline for a more serious approach than the European one: Shuting eyes in murky waters and telling everybody afterwards how clear they are.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
ElMaestro
★★★

Denmark,
2010-04-14 21:31
(5096 d 19:08 ago)

@ Helmut
Posting: # 5140
Views: 9,683
 

 Outlier detection in BE

Dear HS,

❝ • One has to know that there are some (i.e., not look at the data

  driven by some gut-feeling - but perform statistics


Well... I am not so sure about that. Lemme ask a simple question: What is an outlier?
To me, an outlier is a point whose presence cannot credibly be accounted for by random scatter or natural variation, but whose presence has another and identified explanation.
To me, an outlier is not just a point that somehow looks, smells and feels different from the other points.

I am not a big fan of the practice of running some sort of outlier detection test and then, solely on basis of statistics, deeming this or that point an outlier.
Do slow metabolisers cause outliers, if their phenotype occurs in say 0.1% of the population? No, because the data points they cause represent natural but rare variation.
Do exploding chromatolophystic columns cause outliers? Yes, because the peak is not on the harddisk but sitting on the lab technician's glasses.

Here and there one can observe the odd data point space that is just not a credible (credible in this sentence is not quantitative) representation of a model and some scatter and where there really is no identified source of the issue. That's not an outright outlier, but a matter for an auditor or inspector to look into, and it should in my opinion remain so unless a plausible reason can be identified.

Best regards
EM.

Pass or fail!
ElMaestro
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2010-04-14 22:00
(5096 d 18:40 ago)

@ ElMaestro
Posting: # 5142
Views: 9,649
 

 Outlier detection in BE

Dear ElMaestro!

❝ What is an outlier?

❝ To me, an outlier is a point whose presence cannot credibly be

❝ accounted for by random scatter or natural variation, but whose presence

❝ has another and identified explanation.


Agree.

❝ To me, an outlier is not just a point that somehow looks, smells and feels

❝ different from the other points.


Agree. That’s why I called for statistics. Nice to sit here in Brussels – the beer looks, smells and tastes different from Vienna’s. No statistics needed here (not outliers, but a different sample).

❝ I am not a big fan of the practice of running some sort of outlier

❝ detection test and then, solely on basis of statistics, deeming this or

❝ that point an outlier.


Did I say that? I said first you have to know that there is one – based on some objective method of detection – and then look for a cause.

❝ Do slow metabolisers cause outliers, if their phenotype occurs in say 0.1%

❝ of the population? No, because the data points they cause represent natural

❝ but rare variation.


Sure. There’s even a term for that type of outlier: concordant outlier – nothing to worry about in BE, but may be of interest (see the end of this post for a surprising example).

❝ Here and there one can observe the odd data point space that is just not a

❝ credible (credible in this sentence is not quantitative) representation of

❝ a model …


Yes. That’s a discordant outlier. Affects BE. All the reasons I have given in my last post may be applicable…

❝ … and some scatter and where there really is no identified source of

❝ the issue.


or the mysterious S × F interaction.

❝ That's not an outright outlier, but a matter for an auditor or inspector

❝ to look into, and it should in my opinion remain so unless a plausible

❝ reason can be identified.


Again: You have to be suspicious first in order to start a search. I would prefer to base my strategy on a pre-defined method rather than saying: Oops, looks, smells and feels strange…

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
ElMaestro
★★★

Denmark,
2010-04-14 23:31
(5096 d 17:09 ago)

@ Helmut
Posting: # 5143
Views: 9,591
 

 Outlier detection in BE

Hi HS,

❝ Did I say that? I said first you have to know that there is one –

❝ based on some objective method of detection – and then look for a

❝ cause.


No I don't think you said that. I think we agree to a large extent :-), although I am perhaps leaning towards having a reason first and then look for the way the reason affected the data. But this is not always possible or feasible in real life.

❝ Yes. That’s a discordant outlier. Affects BE. All the

❝ reasons I have given in my last post may be applicable…


❝ ❝ … and some scatter and where there really is no identified source of

❝ ❝ the issue.


❝ … or the mysterious S × F interaction.


Yes, I am aware of the terminology; my personal preference is to not make a distinction between SxF and D.O. unless a reason for doing so exists.

❝ Again: You have to be suspicious first in order to start a search. I would

❝ prefer to base my strategy on a pre-defined method rather than saying:

❝ Oops, looks, smells and feels strange…


Search for what? Search for aberrant values and then look for causes or look for causes and only then ID outliers? As mentioned above I am leaning towards the latter, but acknowledging there is a real life out there.

During your stay, I hope you'll find pleasure in the pride of our country: the Trappist/Abdij beer. While Leffe, Grimbergen, Westmalle and Chimay turned mainstream a long time ago, I'd recommend a Corsendonk. But be carefil with the amounts or you will wake up tomorrow with pain in your ovaries.




































Sorry dlabes, I could not help it. Forgive me. :-D

Pass or fail!
ElMaestro
d_labes
★★★

Berlin, Germany,
2010-04-15 10:13
(5096 d 06:26 ago)

@ ElMaestro
Posting: # 5146
Views: 9,561
 

 Men - Women?

Hi Guys,

❝ ... I'd recommend a Corsendonk. But be

❝ carefil with the amounts or you will wake up tomorrow with pain in your

ovaries.


Seems some men on this forum are realiter (translates to "in fact") NLYW! :cool:
Or outlier :-D?

Regards,

Detlew
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,638 registered users;
65 visitors (0 registered, 65 guests [including 5 identified bots]).
Forum time: 15:40 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