If you have multiple concentrations within each drug, you could use a factorial design approach with interactions between drugs and concentrations.

You would then need to run post-hoc tests to determine which of the drugs (and concentrations) were effective. Because of the multiple comparisons, you will need to adjust the estimated p values to account for multiple testing (such as Bonferroni, etc.).

Thank you Ariel. But I am still confused about the type of statistical analysis I should use. I know nothing about factorial design approach. I am not a statistician basically so I can only use Already available softwares and programmes. Can you please help with this.

Sonal,

My suggestion to you is to find a local statistician to work with. If this work is important to you, then you need a trained individual to assist you. Statistics is more than pushing buttons on a computer.

Perhaps, you may find individuals here on RG that may be willing to help you with your analyses.

Ariel

 will you give me some details of data to suggest solution further as with present scenario it will be difficult for every one to suggest solution.

Regards

Rafique

Sonal, 

The principle factor determine the experimental design is the uniformity of the experimental units ( animal , plant , ...... ) which you used in the experiment,  if your experimental units uniform then you can chose factorial experiment conduct with completely randomized design ( of course you must do replicates for each treatment ) , if your experimental units have one variation factor ( different age , class , breed ,.... ) you can chose factorial experiment conduct with Randomized Complete Blocks Design .

Number of replicates effective factor on accuracy of experiment .

Dear Sonal,

Ariel and Mario gave good answers.

If you are confused then I also strongly recommend the colaboration with a local statistician. It will also give you the opportunity to learn a lot (especially about science, because statistics is the foundation of all quantitative empirical science!).

There is no simple "do-this-and-be-happy" solution. One must think (hard) how to analyse the data so that the results of the analysis will be meaningful and interpretable and make efficient use of the information in your data.

This includes to understand what kind and amount of information is potentially available from your data (-> what is the kind of data you wish to analyze), and to know the details of the experimental protocol. Then one needs to translate biological hypotheses into statistical hypotheses. For instance: it is relevant to know if you hypothesize that the drugs work on the same pathway, independently, competitively, or synergistically. This would have consequences in how to set up a model that most efficiently adresses this hypothesis statistically (like an interaction model or Bliss independence model etc). Forther you have cell culture data. Here is an important issue the level of replication. Typically cell-culture experiments comprise several layers of replication (primary cultures / passages / dishes / measurements). Replications share common information and the statistical model correctly handle this. A typical way is to use hierarichal (multilevel or mixed) models.

The statistical analysis is an art, and it needs a lot of understanding, knowledge, and experience to do it well. This should not discourage you. It should rather encourage you, short-term to collaborate with a statistician and long-term to better understand statistics and find out how valuable it is to "learn thinking" (in a scientific way), and how much it promotes to conduct better science.