Probit analysis for LD50 calculation

[Charlie]

I am trying to do probit analysis for 95% confidence Limit and it is my first time to use this procedure.

I've got an example in .xls file which is downloaded from https://probitanalysis.wordpress.com/2016/07/07/first-blog-post/.

I used codes as following (reference to LD50 calculation for mixture of dose concentrations in community😞

data A;
input dose total dead;
cards;
0.03 10 0
0.10 10 1
0.30 10 5
3.00 10 8
;

proc probit data=A log10 plot=predpplot;
model Dead / Total = dose / d=logistic InverseCL(prob=0.1 to 0.9 by 0.1);
run;

However, the results of 95% Fiducial Limits are different from example and proc probit dose not provide information for curve fitting ( I need to use proc reg?). How can I reproduce the result from the example file?

Furthermore, I used the same code for my data;

data B;
input dose total dead;
cards;
0.64 12 9
0.8 12 10
1 12 8
1.25 12 9
1.56 12 9
1.95 12 8
;

proc probit data=B log10 plot=predpplot;
model dead / total = dose / d=logistic  InverseCL(prob=0.1 to 0.9 by 0.1);
run;

The 95% Fiducial Limits didnot estimate in the result and there was no warring from SAS. I am wondering if I could find any suggestion or solution for my data.

[PGStats · Accepted solution]

Look at the (traditional graphics) graph produced by :

data B;
input dose total dead;
cards;
0.64 12 9
0.8 12 10
1 12 8
1.25 12 9
1.56 12 9
1.95 12 8
;

proc probit data=B log10 plot=none;
model dead / total = dose / d=logistic  InverseCL(prob=0.1 to 0.9 by 0.1);
predpplot vref=0.5;
run;

LD50 is so far from the data range and the confidence interval so large that they would be meaningless.

[Charlie]

Hi, PG

Thank you for your reply, it really resolves my doubt.

Charlie