Comparing mortality/survival rate among treatments

[#1]

What is the best method to analyze percent mortality or percent survival (100-mortality) data? I am providing a mock data set. I place 20 animals in a cage and at the end of the study I count the number of live animals and the difference from 20 is the dead animals. The experimental design is a complete randomize block design and I want to use block as either a random effect, covariable, or other suggestion. Time of death is no important and thus I'm struggling to use the JMP survival tools. I want to know if there is a statistical difference in mortality or survival rate (as a percent) among the 3 treatments. If I calculate by hand survival_% is: Control=97%, Drug=88%, Placebo=79.5%; and thus mortality_% is: Control=3%, Drug=12%, Placebo=20.5%.

Please provide step-by-step instructions (e.g., Analyze > Fit Model > etc.) or pictures.

Thanks,

Cage Block Treatment Initial_count Final_count Survival_% Deaths Mortality_%
1 1 Drug 20 20 100 0 0
2 1 Control 20 19 95 1 5
3 1 Placebo 20 16 80 4 20
4 2 Placebo 20 17 85 3 15
5 2 Drug 20 16 80 4 20
6 2 Control 20 20 100 0 0
7 3 Control 20 19 95 1 5
8 3 Placebo 20 15 75 5 25
9 3 Drug 20 18 90 2 10
10 4 Placebo 20 14 70 6 30
11 4 Control 20 20 100 0 0
12 4 Drug 20 19 95 1 5
13 5 Placebo 20 15 75 5 25
14 5 Control 20 20 100 0 0
15 5 Drug 20 17 85 3 15
16 6 Drug 20 18 90 2 10
17 6 Placebo 20 17 85 3 15
18 6 Control 20 20 100 0 0
19 7 Placebo 20 16 80 4 20
20 7 Drug 20 18 90 2 10
21 7 Control 20 20 100 0 0
22 8 Drug 20 17 85 3 15
23 8 Placebo 20 16 80 4 20
24 8 Control 20 19 95 1 5
25 9 Control 20 18 90 2 10
26 9 Drug 20 17 85 3 15
27 9 Placebo 20 16 80 4 20
28 10 Placebo 20 17 85 3 15
29 10 Drug 20 16 80 4 20
30 10 Control 20 19 95 1 5

[#2]

This case is a classic application of 'probit analysis.'

1. Select Analyze > Fit Model.
2. Select Initial Count and Final Count (in that order), then click Y.
3. Select the other data columns that you want to use as factors and click Add.
4. Click Personality and select Generalized Linear Model.
5. Click Distribution and select Binomial.
6. Click Link and select Probit.
7. Click Run.

That should do it.

I have to say that this process is backward: collect the data and then decide on the analysis. The choice of the analysis should dictate the data collection.

Hope that these instructions help!

[#3]

@Mark_Bailey thank you for the suggestion. Unfortunately it does not work, I have tried the binomial distribution with not successful results; it does not accept a random variable or covariable. Please try to use the mock data set I provided to see if you can find a suggestion. Cheers,

[#4]

You are correct. The GLM platform does not accept terms for random effects. The JMP Pro Mixed Model platform does not accept non-normal error distributions.