4 replies to this topic

[AgeVivo]

Hello

The reported survival experiment used only n=6 rats per group. While writing the LongeCity critique (http://www.longecity...lates/c60health) about the article (http://extremelongev...0-Fullerene.pdf), I had a debate with the LongeCity reviewer about whether the differences in lifespan are statistically sufficient.

To me it is, as I describe in the next posts. But I am inclined to be demonstrated differently.

Best,
AgeVivo

Edited by AgeVivo, 15 July 2012 - 12:28 PM.

[AgeVivo]

I find that the reported differences in lifespans are very statistically significant.

This morning I have reproduced it with an online tool so that you can double-check if you wish:
go to http://iscc-serv2.im...er/K-M_plot.php
copy the following in the text box:
"7,1,1
15,1,1
17,1,1
19,1,1
22,1,1
23,1,1
22,1,2
23,1,2
25,1,2
26,1,2
40,1,2
40,1,2
"
I deduced this data from the corrected survival graph; each line is an
animal; the first number is the lifespan (in months) since treatment start;
the second number means that the animal is dead rather than "lost"; the last
number is the treatment group.
click on "addData", then scroll down. you get a very low p-value "p= 0.0046"
: this is below 0.05 so in common terms the difference between water-treated
and olive-oil treated groups is significant and this is even below 0.01 so
it is very significant

click on "clearAdded" and change the data as you want to do other tests.
here is the data I noted for the 3rd group:
"41,1,3
42,1,3
42,1,3
43,1,3
44,1,3
44,1,3
"
I get the the difference between olive-oil and c60oo is very statistically
significant (p=0.0008). And of course the difference between water and c60 is
even more statistically significant (p=0.0005)

You can even slightly modify the data, you will still get very low p-values:
so to me, empirically, the result is very "robust". Note: I am not used to discuss
"statistical robustness" (http://en.wikipedia....bust_statistics) so it might have a different meaning.

Of note: Usually lifespan differences are much more subtle than in
this experiment, therefore the need for many animals (here is a table I have
done here: http://agevivo.com/stats/LOLES.html; it can be used for B6 mice
as well as wistar rats).

Here is more /// more that I guess will be accepted by some statisticians
and refuted by others (because it sounds extremely strange at first): in
parametric tests, we implicitly add knowledge we believe in. When almost
doubling lifespans, n=1 can even be sufficient to state that the lifespan
increase is statistically very significant, compared to the many other
lifespan tests that have been performed with the same species in various
conditions. Roy Walford used to say it as such: "show me the 45 month old
mouse". Therefore my n=3 experiment at home may lead to very statistical
results in some statistical sense.

[AgeVivo]

In case you don't trust the logrank test and in case you know R (very respected freeware, available at http://www.r-project.org/), here is the p-value between olive-oil and C60oo groups computed differently:

library(survival)
survdiff(Surv(c(41,42, 42, 43, 44, 44, 22, 23, 25, 26, 40, 40), (1:12)*0+1) ~ c((1:6)*0+3,(1:6)*0+2), rho=0) #we get p=0.0008
survdiff(Surv(c(41,42, 42, 43, 44, 44, 22, 23, 25, 26, 40, 40), (1:12)*0+1) ~ c((1:6)*0+3,(1:6)*0+2), rho=1) #we get p=0.0015

We still get very low p-values indicating that the differences in reported lifespans between olive-oil and C60+olive oil are very statistically significant.

Edited by AgeVivo, 15 July 2012 - 12:12 PM.

[Turnbuckle]

I deduced this data from the corrected survival graph; each line is an
animal...

Probably not, as there aren't enough for six animals in each group. Probably more than one animal died in one month and thus are shown together.

[AgeVivo]

Probably not, as there aren't enough for six animals in each group. Probably more than one animal died in one month and thus are shown together.

I meant that "7,1,1" corresponds to the rat that died at 10+7=17 months in the 1st group (water treated), "15,1,1" to the rato that died at 10+15=25 months in the 1st group, etc. In total there are well 6*3 = 18 lines. Do you see a mistake somewhere?