Skip to content

T50 or ED50 for count data

4 messages · Mehdi Abedi, Philippi, Tom

#
Dear Colleagues,

I am interested to calculate time to 50 percentage for count data.
I am working on soil seed bank germination. I have recorded number of
seedling emergence in 10 different times for species. The total number of
seeds in soil is not known therefore it is not proportional data but we can
argue that after this period all viable seeds already germinated(total
seedling emergence).

I am familiar with drc package for calculation T50 but only for
proportional data like germination in petridishes which total number of
seed in each pteridishes is known  T50 could be calculated.

Could we calculate T50 for count data as well? because we have total
seedling emergence in the end, therefore their distribution during time is
important and having this index we can find when species germinate ...
This indices could be useful for other count data specially survival
studies.

warm regards
Mehdi


*Mehdi Abedi Department of Range Management*

*Faculty of Natural Resources & Marine Sciences *

*Tarbiat Modares University (TMU) *

*46417-76489, Noor*

*Mazandaran, IRAN *

*mehdi.abedi at modares.ac.ir <Mehdi.abedi at modares.ac.ir>*

*Homepage
<http://www.modares.ac.ir/en/Schools/nat/Academic_Staff/~mehdi.abedi>*

*Tel: +98-122-6253101 *

*Fax: +98-122-6253499*
#
Mehdi--
Based on the example datasets germination and chickweed, drc would work
with you using the final count of emergence as the number of seeds at risk
of germinating.  If you get a satisfactory fit from one of the 6 or so
function forms it supports for dose-response, you may be fine.  [I have
scientific qualms about treating time as a dose, but you may have a valid
justification.]

However, I don't know that drc handles replicate subjects (petri dishes)
followed over time, where you need a random effect for subject.  [You don't
say whether you have replicate soil samples followed over time.]  If you
have replicate soil samples,  you could fit something like a generalized
linear mixed model with lme4 or related, or generalized additive mixed
model with mgcv or gamm4. family='binomial', and for each subject &
species, the sum of the seeds that germinated as the total.
Alternatively, there are a number of packages for survival or time to event
analysis.  Your T50 is equivalent to the median survival time (or median
time to germination).  Look at the survival analysis task view:
*https://cran.r-project.org/web/views/Survival.html
<https://cran.r-project.org/web/views/Survival.html>*
You may be able to use the survival package, as median survival times are
available in most packages.  Gordon Fox (currently at USF) wrote a book
chapter on applying such methods to germination times (albeit using SAS not
R at the time), and might be willing to send you a pdf.
I hope that this helps you with your analyses.

Tom 2
On Thu, Jul 23, 2015 at 2:03 PM, Mehdi Abedi <abedimail at gmail.com> wrote:

            

  
    
#
Dear Tom,
Thanks for your kind reply.
Could we calculate T50 in gamm4 and lme4 as well?
I searched in survival but no special information about median term in the
package survival.

Warm regards,
Mehdi
On Fri, Jul 24, 2015 at 4:11 AM, Philippi, Tom <tom_philippi at nps.gov> wrote:

            

  
    
#
Mehdi--
You _can_ calculate estimates of the parameters at median germination from
the lme or gamm objects, but the confidence intervals are not easy.
I recommend using the survival package, specifically survfit.formula and
quantile.survfit, which gives a couple of examples

fit <- survfit(Surv(time, status) ~ ph.ecog, data=lung)

quantile(fit)

Tom 2
On Fri, Jul 24, 2015 at 12:43 AM, Mehdi Abedi <abedimail at gmail.com> wrote: