I have been trying to do this for a while now, but does anyone know if it’s really possible to calculate the mean lifespan using graphpad prism 5.0 software? I choose the Survival chart type, but how do I put values from two independent experiments (e.g. one experiment done twice) for the same strain over a period of time? I start with a synchronized population of 1 day-old worms (100% alive) and count number of worms dying every day until the last worm dies (0% alive).
Since I couldn’t figure out a way to put results of multiple experiments, I added up values for each strain of the same day and inserted it in the column corresponding to that day. This did give me an expected curve and a median lifespan, but is this method really sensible…? Between mean and median lifespan values, anyway, which one seems more preferable to show?
Thank you very much for your generous help. Cheers.
I’ve never used GraphPad, but recently used the OASIS (online application for survival analysis) which was described in a 2009 PLOS ONE paper. I found that it was pretty straightforward and worked well. If you’re interested, here’s a link: http://sbi.postech.ac.kr/oasis/introduction/
One can upload or paste in data from excel, the instructions on the site are fairly clear and one can get it working without too much trouble. The program then generates survival plots and gives Kaplan-Meier statistics, and mean and median lifespans. With the OASIS input, one would have to combine the results of independent experiments, as you’ve done, and upload the cumulative data.
Thanks, Jordan. Did you need to install Python too? I went to the webpage and installed source codes for survival analysis off the Oasis and installed the Python too…can’t understand where the actual Oasis software interface is or when I’m supposed to input values…
You can paste or upload the lifespan data from this page. Instructions on the format required for your lifespan data are under the “Instructions” link. The web interface worked great and generated all the statistics that I required and was pretty straightforward.
So you’re trying to pool your data? Or are you trying to compare the two independent trials? Either way, I put how to do both in a Prism File. You can grab it from my DropBox:
Hi Snug, thank you for your response. Actually I was able to draw curves. My problem is in getting the mean of two trials. Suppose I do a lifespan measurement twice, and so have two sets of data for the same strain.
How do I obtain the mean lifespan of these two trials?
Unfortunately that’s the one limitation with GraphPad. You can’t do that (or at least I haven’t figured it out after reading the manual many times), mainly because of the way you’re entering the data (all those 1’s…). I find the easiest way to do it is use Excel or another spreadsheet program.
Copy your GraphPad data over into a spreadsheet so that you’ll have two columns, one for ‘days’ and one if your animal died or not. In the column next to it, assuming you have headers, make the C2 cell = A2*B2. Fill the formula down for however many samples you have. Then, go to D1 and just put in the formula =AVERAGE(C:C). Kind of a round about way to do it, but that’s life.
i am doing 4 replicate plates for my survival assay.
sometimes i do get some cases where, there are much worms being censored away,
but in the end of day, when you calculate for the mean and s.d. value on each time point (using excel),
the data still makes sense for comparison.
however, i found out during kaplan-meier calculation, i am actually pooling all the replicate data together,
with the worms censored etc issue, it actually puts a weight to the final calculation
(i.e.: taking an average of 4 replicates Versus taking an average as a whole, they are two different things)
if that’s the case, which one should i believe in then? ???
Hmm, he’s probably getting lots of bagging and needs to censor. Maybe in your case you could report the individual statistics for each experiment? How much weight does the censoring put on?
OK…just going back to the question regarding getting the mean for two trials. I use R for my stats and not Graphpad Prism, but isn’t it possible to use the subcolumn stats function in GPP to calculate the mean of your two trials?
Also, in the case of ‘pooling’ 4 sets of data rather than assessing them separately, I guess it depends on the stats that you do on your data. If GPP assumes that you are doing multiple testing, then the conditions for significance will be more stringent than they would for a ‘single’ test. Hence the importance of the amount of censored data in your trials.
Although you are doing 4 tests, I assume that the worms all come from the same plates/source and therefore are effectively one sample? hence one could say that you are testing a large population split in 4. Here you could pool your data. Or is it the case that the 4 trials took place over a period of time and using different plates/sources? In which case, you cannot pool the data.
Hmm, I think for the column stats function you still need to input the individuals as for example, “16” rather than as a “1” (which would correspond to a 16) if an individual died on day 16.
I personally use R or Excel for everything too (GraphPad is just to make those pretty graphs…).
yes, i am using GraphPad Prism 5 for kaplan-meier analysis. (getting version 6 soon)
i censored the worm as they crawled up to the plate wall (‘walling’).
i am doing host-pathogen interaction study (thus the survival assay comes). the censor did gave impact to the overall graph, if the censored worms are way too much.
i agree with steveh’s saying about “Although you are doing 4 tests, I assume that the worms all come from the same plates/source and therefore are effectively one sample? hence one could say that you are testing a large population split in 4”
indeed it’s true. the worm derived from the same plate source, so they are technical replicates from a large population.
Ok, then aside from the issue of censoring the worms climbing out, you have effectively one large sample? hence you enter the data as one large sample and get the mean for one large sample…
or am I missing something? If this is the case, then the calcs become relatively simple and you’re without a dilemma, believe the combined test (always good on a Thursday and heading towards the weekend)
oh yes, you probably have, but have you tried copper rings to keep them in? Or perhaps (I’ve read this somewhere) silane on the well walls, then they can’t get a grip…
