In my study, I will perform embryo bisection. The blastocyst rate is calculated by the ratio: (blastocyst in seven days / No. of oocytes that were used in IVF) * 100 , and then, I do the comparisons among treatments. However, due to bisections the number of blastocysts can be greater than number of oocytes, that is, blastocyst rate can be greater than 100%. I have read many statistical methods (logistic regression, ANOVA with data transformation, chi-square) but none offered to me a reliable solution.
Can anybody indicate to me a statistical method?
Well, maybe there is a bad deffinition of the blastocyst rate. ~_^
Anyway, try to use contingency tables with defined treatment and number (no percentatge, but count) of blystocysts and oocytes. It should work.
(by the way, it does not seem to me you have data for either ANOVA or regression)
Dear Frantisek, thank you for answering my question.
I was trying to solve that problem and I think I got a solution. In my study I'll do 4 replications; so, considering the blastocyst rate for each treatment in each replication I 'd have four blastocyst rate per treatment.
My problem is how to proceed if some of these blastocyst rate exceed 100%, my solution is to compare not the percentage but the ratio (i.e. 90% = 0.9). If my blastocyst rates do not exceed 100%, I can simply transform data (arc sin transformation); however, if some of blastocyst rates exceed 100% I can do a square root transformation (x'= √(x+0,5) using not the percentage but the ratio (i.e 110% = 1.1) and then to perform ANOVA.
However, I´d like to know if anyone has another suggestion.
I am not sure, if different data transformation from some constant (e.g. 100) is a good idea. It is the same as if you added 10 to one group, multiply the other one and than wondered about differences.
So, you have 4 replications for a treatment/group? If so, there are 4 values per group? I am afraid it is too little for ANOVA.
If the only one goal of yours is to show where is more blastocysts than oocytes, do no statistics, the ratio is fine then. If you are going to show where the difference between the group and 100 is significantly higher (i.e. there is significantly more blastocysts than oocytes), use Student's t-test in these groups. If you want to compare the treatments, use contyngency tables.
I think it is not the blastocyst rate that has to be assessed. The only thing is that you should have first the success in bisect ion (bisected vs total) and then total blastocysts got after no bisected successfully. May be by this the statistics can infer a good prediction or conclusion.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude