# Dont know which statistical test I should use?

#### MDRLOz

##### New Member
Trying to see if a relationship between glucose concentration and cell death.
I am measuring the % population of dead cells on a time course (96 hours) post drug treatment (hour 0).
I measure the dead cell population every 24 hours giving me 5 data points.
I also measure the glucose concentration (starts at 12mM) at each time point.
If that is not enough data I also have the same experiment done again but starting at 50% glucose (starts at 6mM) and I have completed all the above on 6 different cell cultures.

#### staassis

##### Active Member
Simplest methods of survival analysis, like Kaplan-Meier estimator or Cox model, would not be applicable here. The reason is we have a continuous predictor (glucose concentration) which changes randomly over time. The survival of each cell is best characterized by a counting process with stochastic intensity of the form

λ(t) = f(t, Glucose(t), ...).

In particular, you can try the following specification:

λ(t) = exp( B0 + B1 * t + B2 * t^2 + B3 * Glucose(t) + ... ).

The model can be estimated using the method of maximum likelihood. As a byproduct, you will get the p-value of Wald test for each parameter... Arguably, the easiest path is to program this method in R or Matlab.

#### staassis

##### Active Member
Or

λ(t) = exp( B0 + B1 * t + B2 * t^2 + B3 * Glucose(t) + B4 * log(Glucose(t)) + ... ).

#### MDRLOz

##### New Member
Thanks for the infp. Not sure I am up to that level. Was more looking for that I needed to do two-way Anova or something.
However, I will take what you have given me and ask a stats person for some help.