agreement analysis?

[Ashish Awasthi]

Greetings from India

I want to check repeatability of a medical test, for this purpose I have data (continuous) on two time points  for each individual 

I am confused which method I use?

Bland Altman mean difference for agreement analysis or Concordance correlation coefficient , or some other method

Thanks

--
    Regards
Ashish Awasthi
PhD Scholar

Dept of Biostatistics & Health Informatics
SGPGIMS, Lucknow-226014

(M) 9208604604

[Frank Harrell]

The simplest approach is to compute the mean (over observers at the same time) absolute difference between two measurements.  This approach is described in http://biostat.mc.vanderbilt.edu/twiki/pub/Main/ClinStat/obsVar.pdf

Frank

[Rajeev Kumar]

Hello Ashish

You can try Intraclass correlation for absolute agreement by using two-way mixed ANOVA.

Rajeev

--
--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules
 
---
You received this message because you are subscribed to the Google Groups "MedStats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to medstats+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

--
Rajeev Kumar
Department of Biostatistics and Medical Informatics
University College of Medical Sciences
Dilshad Garden
Delhi - 110095

[BXC (Bendix Carstensen)]

The intra-class correlation depend on two things: 1) The variation between measurements at the two time points (WITHIN person variation) and 2) the variation between the patients’ measured levels (BETWEEN person variability).

 

Now if you remove some  persons with values in the middle of the distribution, you will increase the BETWEEN person variability, but not influence the WITHIN variability much. The result being that the ICC increases, but clearly the repeatability (WITHIN person variation) remains the same.

 

So what does the ICC tell you in those two circumstances?

 

In my opinion ICC should be avoided in this type of circumstances. The only two advantages are 1) that it is easy calculate if you don’t bother to consider what you really want, and 2) it will paralyze the very large proportion of the medical community that (understandably) do not have a detailed knowledge of the concepts.

 

b.r.

Bendix Carstensen

[Ashish Awasthi]

I also do not want to use ICC because its not easy to understand by non Statisticians

I am confused over two methods which I am planning to use , 

Bland Altman Agreement Analysis available at 

www.medicine.mcgill.ca/epidemiology/hanley/tmp/surveys/ba.pdf

or

concordance correlation available at 

http://www.jstor.org/stable/2532051

both of these measure agreement between two consecutive measurement of same test but question arises which one is better


Ashish

[BXC (Bendix Carstensen)]

Concordance Correlation has exactly the same drawbacks as the correlation. And so should never be used in this type of studies.

Use the Bland-Altman analysis if you have a clear first and second measurement. If your two measurements are exchangeable, use the method that Frank Harrell gave a link to.

b.r.

bendix

[Basilio de Braganca Pereira]

Have a look on
LUIZ, R. R.; COSTA, A. J. L.; KALE, P. L.; WERNECK, G. L. Assessment of
agreement of a quantitative variable: a new graphical approach. J. Clin. Epidemiol.,
New York, v.56, n.10, p.963-967, 2003.
LLORCA, J.; DELGADO-RODR¶IGUEZ, M. Survival analytical techniques were
used to assess agreement of a quantitative variable. J. Clin. Epidemiol., New York,
v.58, n.3, p.314-315, 2005.
Basilio

2013/5/5 Ashish Awasthi <ashish...@gmail.com>

[Raed Bahelah]

I believe you should give more details about what you exactly want to do: Do you want to test the difference over time in comparison to the baseline or something else?

What Mr. Frank Harrell refers to is the absoluet difference from baseline which isn't much helpful as it doesn't control for covariates and it assumes Zero variability between subjects at baseline, but you can still use it if that's what you want.

If you want to study the intraobserver agreement, then that's something different.

So, more details are needed.

 

[Frank Harrell]

I don't understand what I said that made you interpret it as being interested in change from baseline.

Frank

To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.
 
 

 

--
Rajeev Kumar
Department of Biostatistics and Medical Informatics
University College of Medical Sciences
Dilshad Garden
Delhi - 110095

--
--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules
 
---
You received this message because you are subscribed to the Google Groups "MedStats" group.

To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.
 
 

--
--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules
 
---
You received this message because you are subscribed to the Google Groups "MedStats" group.

To unsubscribe from this group and stop receiving emails from it, send an email to medstats+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.
 
 

--
    Regards
Ashish Awasthi
PhD Scholar

Dept of Biostatistics & Health Informatics
SGPGIMS, Lucknow-226014

(M) 9208604604

 

--
--
To post a new thread to MedStats, send email to Med...@googlegroups.com .

MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules
 
---
You received this message because you are subscribed to the Google Groups "MedStats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to medstats+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Raed Bahelah]

Prof.Frank Harrell:

"absolute difference between two measurements", it seems as you talk about change from baseline combined with what posted in the original question which had the following phrase "..on two time points for each individual", so I thought you mean what I said.

[SadaNand Dwivedi]

I agree with Frank, it may often serve the purpose especially if one has clarity about clinical relevance of calculated mean difference. However, for more clearer picture, it is better to calculate all measures -starting from correlation coefficient, to intraclass correlation, regression coefficient in regression line without intercept; confidence interval for mean difference & then take complete view before commenting on agreement.

 

Sada Nand

 

S.N. Dwivedi, Ph.D., FSMS

Professor, Department of Biostatistics

All India Institute of Medical Sciences

Ansari Nagar

New Delhi-110029, India

Tel: 91-11-26588441 (Direct-Residence)

      91-11-26593394 (Residence)

      91-11-26593387 (Office)

      91-9810571956 (Mobile)

Other Emails: dwiv...@hotmail.com

                     dwi...@aiims.ac.in

                     dwiv...@yahoo.com

[Ashish Awasthi]

Thanks a lot to every one for their valuable suggestions

Ashish Awasthi
(From Nokia Mobile)

[Pedro Emmanuel Alvarenga Americano do Brasil]

StatMasters,

It sounds to me a reliability study of a diagnostic test. There were several good comments so far. 

Just adding a few things: Even for those who are very familiar with diagnostic test development, the reliability studies are always debatable as this is one point seldom discussed in text books. Nevetheless, there are some good references about this issue. 

There is a famous Sklo's book "Epidemiology beyond the basics" which has a whole section about this topic. 

Basilio provided two references which I like, this technique is not very intuitive at first but it is quite nice. 

Also, I suggest take a look at: 

Gwet KL. Handbook of Inter-Rater Reliability (3rd Edition): The Definitive Guide to Measuring the Extent of Agreement Among Multiple Raters. Advanced Analytics Press; 2012. 

Kottner J, Audigé L, Brorson S, Donner A, Gajewski BJ, Hróbjartsson A, et al. Guidelines for Reporting Reliability and Agreement Studies (GRRAS) were proposed. Journal of Clinical Epidemiology. 2011;64(1):96–106.

Lin L. Corrections: A note on the concordance correlation coefficient. Biometrics. 2000;56(1):324–5. 

I

cant find it right now but I did read last year a systematic reveiew about inter-rater reliability studies in general and the most studies still use  Kappa and ICCC, and worst of all, many of the included studied had inappropriate interpretations about these measures. 

I believe it was published in PloS One, but Im not sure. 

Kind regards,

Dr. Pedro Emmanuel A. A. do Brasil

Curriculum Lattes:  http://lattes.cnpq.br/6597654894290806 
ResearchGate.net: https://www.researchgate.net/profile/Pedro_Brasil2/

Instituto de Pesquisa Clínica Evandro Chagas
Fundação Oswaldo Cruz
Rio de Janeiro - Brasil
Av. Brasil 4365,
CEP 21040-360,
Tel 55 21 3865-9648
email: pedro....@ipec.fiocruz.br
email: emmanue...@gmail.com

---Apoio aos softwares livres
www.zotero.org - gerenciamento de referências bibliográficas.
www.broffice.org ou www.libreoffice.org - textos, planilhas ou apresentações.
www.epidata.dk - entrada de dados.
www.r-project.org - análise de dados.
www.ubuntu.com - sistema operacional

[Frank Harrell]

No I never said change from baseline.  I meant absolute discrepency between two measurements from two different observers at the same time.

I can't think of a context where change from baseline is a really good idea.

Frank