When you treat cells with different drugs (i.e.dose- response; drug alone vs combination, etc..), what kind of biostatistical analysis are better to interpret the results? Sometimes I am not sure if I am doing the appropriate significant test. I have been reading about it, but I am not sure about it.
In that case, I think the suitable statistical analysis will be the chi-square test. It is a categorical x categorical test.
How are you creating your combination studies? There are 2 different valid methods I can think of for this type of study. They are Mixture Designs and Response Surfaces.
When my mom was getting chemo therapy, she was given 3 bags of chemo. The doctor didn't worry about the amount of liquid she was given. He worried about the dose. In this case, he could use an Optimal Response Surface and test the best dose levels of the 3 chemo drugs. If he had a 4th drug to test, he could still use the Optimal Response Surface. He would be able to specify the dose levels of all 4 chemo drugs. He would then use Multiple Linear Regression to analyze the drug doses and their responses.
If the doctor had say a total volume of 2L to deliver all the chemo, then he would use a Mixture Design. In this type of design, each chemo drug would consist of a portion of that 2L. The total volume would always be 2L. To analyze this type of data, you need special stats software.
If you created your own study data for combinations, then you will use Multiple Linear Regression, or something like it.
Should I consider independent or dependent samples if I treat the same cell line with different conditions (i.e. more than 2) and more than 3 independent experiments (different days or weeks). Should I use one-way ANOVA or t- test to establish if the differences are statistically different? There are more appropiadate test? The outcome are used to be numeric (i.e. % of cell death, ATP levels, oxygen consumption).
How many cell lines do you have?
If you have 1 cell line, then there isn't much you can do. If you have 2+ cell lines, you have more of a split plot design. ANOVA and t-tests are not appropriate.
With the independent experiments and different conditions, do you do different combinations of these? If so, this is a factorial design experiment. T-tests and simple ANOVA are not the answer.
For your outcome variables, do you take one reading each, per sample or multiple readings? If it is multiple readings, you have a repeated measures design. Simple ANOVA is not the correct approach.
% cell death, do you get a total number of cells and count every dead cell? If so, this is probably best analyzed as a logistic regression. Your response would be, Ln(P/(1-P)). 'P' is the probability of death. If not, you should probably do Poisson regression.
I have one cell line and I usually do three different experiments (maybe 5 different tratments with different compounds in three different days ) and each experiment with three replicates. So the n=3. Should I consider my samples are dependent or independent. Maybe if they are the same cell lines I should consider it dependent. If I wanted to compare the means of cntl vs treatment 1 and cont vs treatment 2. Neither anova and t test are useful?
Well, if you only have one cell line, you can talk about that one cell line. Not all cell lines. I wouldn't consider your data to be dependent yet.
You can treat "Day" as a blocking factor. Good thing you did this.
How do you define a treatment?
If you mean that you use Drug A, B, C, D and E as the different treatments, That will be one type of analysis. If you use a factorial design, that will be another type of analysis.
What did you use to design your treatments?
If your response variable is numeric, while the independent variables (treatments in your case) are categorical, and if the goal is to identify which treatments are statistically different, you can do analysis of variance. Now, as you have one cell line on which you are repeating measurements, such an anova would be repeated measures anova. Your study sounds like more of a "crossover study" in which same replicate is subjected to different treatments over time. In this case, the observations that are close in time are not independent as the covariances are equal. Repeated measures anova can help you to model these covariances.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis