I am currently doing some studies in which I incubate PBMCs from 4 different donors with a concentration range of a drug for 3 different incubation times.
In other words: PBMCs are purified, 6 different concentrations of the drug (0 / 0.01 / 0.03 / 0.1 / 0.3 / 1 ng/ml) are added and the cells are incubated with the drug for either 3, 24 or 48 hours before I measure cytokine levels.
I would like to test whether the concentration of the drug plays a significant role in the cytokine response and at the same time I would like to know whether the amount of incubation time is a significant factor as well.
Which statistical test should I use? I have considered using a two-way repeated measurement ANOVA but I am not sure whether this makes any sense? I guess both time and concentration would be independent variables whereas the cytokine concentration would be the dependent variable?
I would really appreciate some help, since my statistics skills are really lousy.
Hej Katrine,
Om du använder SPSS så verkar det som om Mixed between-within subjects analysis of variance är en bra metod. Du har tre variabler som är involverade: du har 2 oberoende variabler ( 6 olika doser av drogen och 3 olika tider) och så har du den beroende variabeln cytokinkoncentration.
Okey, jag använder inte det. Two-way repeated measures ANOVA kräver en kategorisk variabel: tid1, tid 2 tid3
Tack så mycket! Jeg vil forsøge at finde en tilsvarende test i SigmaPlot. Har den test du bruger andre navne?
Kära Katrine,
jag använder SPSS survival manual av J. Pallant och här kallas testen så. Skulle dock tänka mig att MANOVA också fungerar i detta fall. Här är en artikel från RG som också har använt detta test. Kolla:
Bjoern,
Is not MANOVA better in that case?
Article A simple method for distinguishing within- versus between-su...
The probelm is that you have not specified your problem statistcally in an unambigious way. "I would like to test whether the concentration of the drug plays a significant role in the cytokine response" is be no means any statistical hyothesis.
To turn this into something useful you would need a (kinetic) model, some assumed functional relationship between dose, time, and response. You will recognize that there is a large variety of possible relatiionships, like for instance a saturation curve, a logistic curve, or even an optimum curve (having some peak) - or yet something entirely different. Depending on this assumed model there will be some parameters having particular meanings. This can be the satuartion (maximum) value, the dose or time for 50% saturation ("EC50"), the initial slope of the curve, the slope at the EC50, but it can also be the dose/time at the peak response, the witdth of the peak, or some measure of the symmetry of the peak etc. So you would need to specify what property of your model you are actually inetested in. Then you can statistically test your model against a model with a resticted vaue for the particular parameter. But even better you would automatically get estimates for the parameters, telling you quantitatively the kinetic properties of your system.
In the model, you should have F(time, dose) = [Cyto].
In general, it sounds like you have a multiple regression model. The type of regression model to use is questionable.
If we decided to use a "textbook" approach, it sounds to me like you are looking at a Poisson Regression. (This is the textbook method for models where the response is a rate/conc.) There are a lot of different considerations for a Poisson regression model though. I also assume that you have say different well plates (multiples of 3), each one is incubated for a different time.
As Jochen stated, there are many different types of models you can use. My copy of Minitab has about a dozen models that might be of use to you. What is the goal of the model?
You should use two way anova, for more information you can visit the link below.
https://www.youtube.com/watch?v=ZAEYgU_kp7I&list=PLtuGiD4XXsJ2SRFAWrrWEqOjBjaL5bAIc&index=9
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology