There were four groups; A, B, C and D with each group containing 15 birds. All birds are of the same age. Other environmental and experimental conditions for all the groups are also same. Group A birds were treated with single dose of drug while, group B birds were treated with double dose of drug treatment. Group C is positive control and group H is negative control. At the end of the experiment, the number of deaths (% mortality) in each group was calculated. I want to know does any significant difference exist in % morality of treatment groups in response to single and double dose of drug?
Hi, Faisal. You can try the Chi-square test followed by its posthoc comparison.
If you are familiar with R software, here are the step:
1-) In excel, create a contingency table consisted of two columns (number of dead and number of alive) and four rows corresponding respectively to the group's A, B, C and D.
2-) Once created, select and copy your contengency contingency table,
3-) open your R software and enter this code: data
Steve B. S. Baleba ,
if you suggest R, why not suggesting a binomial model, like glm(cbind(num_survived, num_dead) ~ group, family=binomial) ?
Depending on what level in group is used as reference, the coefficients will already be the log odds ratios for survival.
If Tukey-contrasts (= all pairwise comparisons) are of interest, one could use
contr
Jochen Wilhelm ,
Your proposition of using Generalised Linear Model with binomial distribution could also assist Faisal Rasheed Anjum
Thank you for the suggestion.
Best.
Faisal Rasheed Anjum, what you are describing sounds like a kind of dose-lethality study (to gauge ld-50) but using only two dosages of a drug? What is the "positive control" that you mention? I don't understand what that could be in your context.
The good "old fashioned" way of doing this would have been to simply plot your empirical results, and create an ad-hoc fit. Then overdose another cohort of birds and see how well your replicated results fit with your first empirical model. The typical approach would have been to use more dosages, too.
Jochen Wilhelm's suggestion is a kind of logistic regression model, if I am not mistaken, which I agree does make a lot of sense if you are intent on a statistical approach.
Ultimately, I think you would be most interested in the contrasts (in mortality) between the two dosages. What is your a priori cut-off for a biologically meaningful mortality difference? Any confidence interval probabilities would be predicated on the distribution curve underlying the model, of which your guess is as good as mine.
You devoted a substantial sample to the control group without the drug, which is good, but do you expect mortality there, as well? What an odd experiment.
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing