We have done an animal trial with two different groups- treated and untreated over a period of 8 weeks, 10 weeks and 12 weeks. Each timepoint, we had two sheep, each containing one treated and one untreated sample. So, each timepoint, we basically had 2 samples for treated and 2 samples for untreated group. Someone suggested that we could take all the 6 data (from all 3 timepoints) together for each of the groups and do T-test to compare between treated and untreated. That's my question. I am not sure if it is the correct way to plot the data. How can we plot all the data from different timepoints together in each group? They have time difference meaning they have differences in biological factors' activity, tissue reactions in response to the implants we used. If that suggestion is true, there is no need to use different timepoints, right?Can we really present different data for different timepoints as one group (treated or untreated)?
Secondly, if we want to compare between treated and untreated, we should do for each of the timepoints, right? In that case, with the given sample size of 2 for each timepoint, is it feasible to get a statistical significance?
I would highly appreciate your expert opinion here. Thank you in advance.
Hello D S Abdullah Al Maruf
"Someone suggested that we could take all the 6 data (from all 3 timepoints) together for each of the groups and do T-test to compare between treated and untreated."
I do not agree with this opinion.
Hi Abolfazl Ghoodjani!!! Thanks for your reply. Could you please explain your answer a little more?
When, very small sample size represents the related population size, then you may use tests.
Öztas Ayhan Thanks Dr Ayhan. I understood what you meant. But, in this case, each timepoint has only 2 data points which we can say as 2 biological replicates. Now, my question is that how can 2 biological replicate from 8 weeks be considered as the same as that of 12 weeks? Samples from 8 weeks should not receive the same amount of time, same tissue response, or stimulation from growth factors compared to the 12 weeks. They are simply different. The only common thing for each of the time point is that they are either treated or untreated. But, the reason we chose different timepoints was also to compare the difference between treated and untreated time-period-wise. It seems confusing to me.
One thing that is certain is that t tests cannot work for small samples (n
Lewis Maccarter Thanks for your explanation. Yes, I agree that the sample size is too small in our case to compare or get a statistical significance. I think that's the reason that person suggested taking all the data for each of the groups so that each group will have at least 6 data points. And, I can't rely on this suggestion as I believe that this will raise questions about why we mix up the data from different timepoints which may not be acceptable at all. Lewis Maccarter Do you think my thought is right?
Yes you are absolutely correct. Your results might be interesting but they are simply two individual case studies. You probably only have information for hypothesis formation. It shows your good sense that you posted this dilemma. I hope your colleague accepts your point. Good luck.
I thought about Bayesian with prior information, but on a second thought, even Bayesian may not be appropriate with the sample size of that small. Perhaps, you can take a mixed study approach (qualitative and quantitative methods) and describe your observation; then you can do one-sample t-test to compare the mean value from the literature. Qualitative observation may strengthen the point you wish to make.
I added the data sheet here. The way he advised is to use one tailed paired T-test taking all the value together as you can see in the spreadsheet.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size