I am doing a time-dose-mortality test on an insect with neem oil. For analysis, I would like to use the Cox regression model from IBM SPSS 20 for Lethal Dose 50 (LD50) determination. I would like to know the interpretation of the beta coefficient (B) and if it is possible to have a coefficient (B) = 17.
Normally the beta-coefficient describes the change in the log-hazard per 1 unit change in your independent variable (IV). Therefore the exp(beta-coef) = HR/1 unit change in IV. However, please provide a bit more information on your experiment.
I guess you exposed insects with neem oil of different dosage for a certain time (number of individuals per dose?) and then you looked if each individual insect was dead or alive. Am I right?
Actually, why are you using a Cox regression? To determine an LD50 I would do an experiment as follows:
10-20 animals exposed per dosage (0, 0.1, 1, 10, 100, 1000 or on the basis of 2?) lets say for 5 days. Afterwards you get the mortality rate for each dosage which you can analyse via a logistic regression with dosage as IV.
After you know the intercept and slope of the regression line (y = 1/[1+exp(-z)] with z = beta0 + beta1*dosage) you insert 50% mortality rate in the equation and solve for dosage --> dosage = (-beta0)/beta1 = LD50.
Thanks for your response
Yes, I exposed insects with 7 neem oil dosages for 15 days and the mortality was daily reported. As you suggest, I will try with Logistic regression (Genstat).
Sometimes with dose-response curves it is better to use the log10(dosage), so I would try both dosage and log(dosage)
To decide if a log-transformation might be better make a scatter plot in advance with mortality rate ~ dosage and mortality rate ~ log(dosage)
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS