I am planning an experiment to check the combination effect of 10 different chemicals for cytotoxicity. I know their individual effects and I would like to check their interaction effect or combination effect on cytotoxicity levels. Some compounds are pure biocides and some have low to moderate cytotoxicity. I came across fractional factorial design of experiments. I am not experienced with statistics so it is quite overwhelming to go through all the basics and calculations. What would be the most convenient way to carry out such an experiment? An expert's advice and help would be much appreciated.
Multivariate methods in tablet formulation suitable for early drug development: Predictive models from a screening design of several linked responses | Request PDF (researchgate.net)
Hi Arif,
to have significant effects of 2nd order interactions you must have a minimum resolution of 4, and for 10 factors you can build a full two-level factorial DOE with a minimum of 32 trials. I remain at your disposal if you want more details.
Hicham CHIBANE
Hi Hanen Hajji, Hicham Chibane
How to choose the level of significance? I have seen various studies where different significance values were used (0.05 was the most used value, however). In order to eliminate the unimportant factors in my study, I need to choose an appropriate level of significance.
Also, while generating a fractional factorial design, how to make sure that we don't confound any significant interaction? For example, If I use the generator function I=ABC, then isn't the effect of ABC confounded?
Regards
Hi Siddharth;
for the factor 0.05, this means that you have an assurance that your model therm is significantly different from zero to 95%. I think it's good to keep 0.05, to stay within the usual norm. And for the fact that I = ABC can be mixed with its alias, you have to check the resolution of your model, to detect with which therm your aliases are mixed.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology