In article <mjcoyne-011293095650 at 134.174.147.40>,
mjcoyne at warren.med.harvard.edu (Michael Coyne) wrote:
> Looking for software for estimating LD50 and LD90 by logit analysis.
> Possible reference, Acharya AS etal., Comput Methods Programs Biomed
> 34(4):255,1991. Any ideas or suggestions?
>> Thanks...
>> Mike
I don't know of software written specifically for this application, but I
can suggest two "work arounds" both of which I've used.
1) You can use a spreadsheet to apply the logit function to your x and y
data (sorry I don't remember the formula for the function off the top of my
head) and then do a linear regression on the results. You can then use the
equation for this line to back-calculate the x values that correspond to
the y values 50 and 90. Disclaimer: I'm not a biostatistician so I can't
say what effect this tranformation has on correlation coefficients, error
estimates, etc., but it definitely works fo calculating the values
themselves.
2) Use a program that does general curve fitting using either preprogrammed
or user-entered formulas. EnzFitter, Sigmaplot, and KaleidaGraph are some
of the ones I know of that do this. After fitting the data you may still
have to back-calculate as descibed above or the program may do it for you.
Hope this helps.
Dirk Bornemeier
Ann Arbor, MI