Appropriate tests for experimental crab growth and survival research

#1
Hi,

I would like to know which test(s) would be most appropriate to analyze my data.

I conducted an experiment with crabs to assess the effects of changing ocean pH on their growth and survival. Crabs were placed in tanks of 3 different pH levels. Each pH level had three tank replicates. Multiple crabs were grown in each of the tanks for 14 days. At days 0, 1, 2, 4, 6, 10 and 14 of the experiments, five crabs from each tank (so 15 per pH level) were removed for measurements (and never returned to the tanks). Therefore, sampling at each time point was conducted in the same tanks but never on the same individual crabs. At the same time points (days 0, 1, 2, 4, 6, 10 and 14), the number of surviving crabs in each tank was counted to assess survival.

1) Growth analysis
So I want to know whether time and pH have an effect on seven different crab size parameters. I therefore have 2 independent variables (time and pH) and 7 dependent variables. Can a MANOVA be used if I have 2 independent variables? Also, how do I deal with tank (3 tanks per pH level) versus crab replication (15 crabs per pH level)? Does this become a nested design?

2) Survival analysis
What is the best way to compare survival over time among the three pH treatments? Is Kaplan-Meier applicable here (with censoring for crabs removed for measurements at each time point)? Or does the lack of independence among crabs grown in the same tank pose a problem?

I have never played with this level of statistics before so any help would be greatly appreciated! Thank you!
 
#2
Generalized Linear Mixed Model for growth and regress growth on time. Even though you're not repeating measurements on each individual crab you still need to use mixed modelling to account for variance within each treatment as well as variance across each treatment; the mixed model approach helps that.

Cox for survival analysis. Since the crabs are in the same tank with pH exposure you can run each crab individually since the Cox regression can factor in the pH category (you can either Dummy variable them or keep them categorical depending on your software)