Talk:Kaplan–Meier estimator
![]() | This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||
|
Confidence Intervals
[edit]Do we need a confidence intervals section? It makes sense to discuss them in relation to Greenwood's formula and maybe introduce a few more methods. Blue down quark (talk) 18:03, 27 March 2018 (UTC)
Error in Table?
[edit]Shouldn't the 8 in the bottom right cell in the first table be a 9? 1 died, so there are 9 at risk prior to t2. — Preceding unsigned comment added by Sprevrha (talk • contribs) 11:34, 21 August 2013 (UTC)
I believe the other unavailable person (making it 8 instead of 9) is subject 4 who was "lost to followup (censored) at day 9". Drevicko (talk) 09:31, 19 July 2015 (UTC)
Copyright vio?
[edit]The image shows 2 curves that are identical to the figure on page 126 of First Aid for the USMLE Step 2 CK, 6th edition, by Le, Bushan, and Skapik. The text labels have been changed and color added, but the curves are exactly the same--even the tick marks correspond perfectly.
—Preceding unsigned comment added by 71.169.146.59 (talk) 01:58, 9 November 2008 (UTC)
gene example is not typical
[edit]The example of gene profiles in the introduction is highly unrealistic. A medical example would typically be the comparison of different treatments for disease Sboehringer 08:53, 13 April 2007 (UTC)
This entry needs more examples, possibly a sample calculation, and a review of the citation history of the original article. I can do it eventually but if anyone else is game, feel free...this is an important entry, so I think a little tutorial on how to do a KM estimation is in order. 63.250.186.17 16:27, 16 November 2005 (UTC)Theoriste
Definition
[edit]I am very unhappy with the definition of the product-limit estimator on this page. On the positive side, the definition uses fairly minimal notation, which is pretty clever (and I assume is the point).
On the negative side, this definition is undefined for t < t(1). That is, the survivor rate is obviously 100% before any subject has died, but the product limit expression is always less than 100% (since N/(N+1) is always less than 1). Perhaps more importantly, this definition has no obvious generalization to the case with right censoring. I much prefer definitions like this one, which I think are more standard.
I would just have edited the page, but I'm lazy, and I'm also very curious about where the definition on this page came from. Unless there are objections, I'm going to go ahead and change it. Ragout 04:50, 15 April 2006 (UTC)
- I think the definition is fine noting that a product with zero terms is by definition defined to take the value of 1. Csaba Szepesvari (talk) 10:01, 24 March 2018 (UTC)
d2_2 ? Costella ref ?
[edit]Surely in the exampled d_2 = 2? —Preceding unsigned comment added by 128.40.38.9 (talk) 14:54, 23 February 2010 (UTC)
Surely the statement that "Data for the first two subjects would be as follows." is incorrect? The current column heading '2' in the table which follows appears to refer to the index i. In total the two numbered columns in the table contains data for the first three subjects.
I don't think the reference to the Costella paper is necessary and suggest removing it.John Lawrence (talk) 22:09, 4 July 2011 (UTC)
- I agree. The reference to the Costella paper is at best orthogonal. I'm removing it. — Preceding unsigned comment added by 98.169.119.68 (talk) 01:45, 16 August 2011 (UTC)
- I third that motion, there are better (text book) resources for the variance of the KM Estimator, I'm not even sure that is the Cramer-Rao asymptotic variance of the estimate, which exists because even though the KM is stated as non-parametric, it has in fact a parameter for ever interval being estimated. — Preceding unsigned comment added by 198.161.2.241 (talk) 17:09, 14 November 2012 (UTC)
Recent long (and possibly tedious) additions to this article
[edit]I think the section should be heavily edited before it is included in the article. --Mcshuffles (talk) 13:21, 4 April 2017 (UTC)
Recently, Michaelg2015 introduced some very long and, I suggest, tedious material to the article: [1]. I propose that this material be removed. Isambard Kingdom (talk) 17:04, 14 January 2017 (UTC)
Isambard Kingdom has proposed deletion of the section "Example calculation of Kaplan-Meier estimate" because it is long (and possibly tedious). I beleive that an example calculation is necessary for a comprehensive description of the Kaplan-Meier estimate. However, I agree that the section is long, and it need not be in the middle of the article; it can be moved to the end for those readers who wish to see the example calculation. I have moved the section to the end. Michaelg2015 (talk) 21:14, 14 January 2017 (UTC)
- We might ask ourselves, who will actually read an article on the Kaplan–Meier estimator? I don't think the relevant reader needs all the tedious detail that is given at the end of this article. Isambard Kingdom (talk) 15:31, 18 January 2017 (UTC)
What is a simple summary
[edit]Is there a simple way to understand the estimator ? Is there a value for each survival curve (as the formula seems to imply) or is it a value that compares two curves ? Does it relate [often] to the area under the curve(s) or the area between the curves ? Is the key point about KM (that could go in the first sentence of the lead) that it notably handles censored (missing?) data. What is the equivalent when there is no censoring to worry about ? - Rod57 (talk) 08:07, 18 July 2018 (UTC)
Need Discussion on Unbiasedness, Efficiency, Consistency, Heteroscedasticity,Autocorrelation robustness
[edit]Needs section on Unbiasedness, Efficiency, Consistency, Heteroscedasticity, Autocorrelation robustness 68.134.243.51 (talk) 14:59, 29 August 2022 (UTC)