I have 6 molecules to evaluate the anti-inflammatory properties and the possibility to inhibits cancer cells. My model is macrofagos (RAW) stimulated with LPS, and I will evaluate ILs, NFkB & iNOS. But I don't have complete security to use a proposed n of 50000 cell by triplicate in each molecule; second I propose to evaluate with ANOVA two-tails, but in a test I find Schef's test but don't have enough weight to pharmacology. Can anybody help me?
I assume your response (outcome) will be amount of ILs, NFkN and iNOS expressed in cells exposed to your 6 molecules. How will the cells be grouped? In other words, are they grown in a medium in a contained unit (petri dish)? If so your dishes are your treatment units. It would be good to have replicates of each molecule and a "placebo" (dishes that are not treated with anti-inflammatory molecule). This means you would have 7 "treatments". If you run triplicates, which is always a good start when you do not know what to expect, then you will need 21 units.
ANOVA is a good way to get a handle on whether or not you have a signal. Each molecule's response outcome can be compared to the untreated levels. You can then order them by amount of response and compared across expressed proteins to see which give the greatest anti-inflammatory response. Usually, these responses are evaluated as "folds"...2-fold expression, etc.
Good Luck!
I assume your response (outcome) will be amount of ILs, NFkN and iNOS expressed in cells exposed to your 6 molecules. How will the cells be grouped? In other words, are they grown in a medium in a contained unit (petri dish)? If so your dishes are your treatment units. It would be good to have replicates of each molecule and a "placebo" (dishes that are not treated with anti-inflammatory molecule). This means you would have 7 "treatments". If you run triplicates, which is always a good start when you do not know what to expect, then you will need 21 units.
ANOVA is a good way to get a handle on whether or not you have a signal. Each molecule's response outcome can be compared to the untreated levels. You can then order them by amount of response and compared across expressed proteins to see which give the greatest anti-inflammatory response. Usually, these responses are evaluated as "folds"...2-fold expression, etc.
Good Luck!
Check out GPower: http://www.gpower.hhu.de/. This is a great tool to calculate statistical power and sample sizes for a range of different analyses and experimental designs.
It is an interesting question, but we usually do not need to worry about the sample size in this situation if you do standard pharmacological/pharmaceutical experiments. "Time course" and "dose course" are often conducted firstly, repeats (one or two more times) are needed regardless statistical tests or other reasons, then you can work out the sample size at certain dose/time points if required according to these data, for example, tools recommended by Pater Klaren. Good luck!
JinBiao Chen has a good point which I failed to mention. You do not really need to worry about sample size at this point. It is more important to have a well-control experiment with replication so that you may get good estimates of the fold changes in protein expression. With that information you can then calculate a sample size that is powerful enough to detect the protein expressions you believe are relevant.
The sample size that is statistically enough depends on the variation of drug effects, the number of groups you have, the expected differences of the effects, the statistical analysis you used, etc. GPower is a good software to determine your sample size if you have some preliminary data in hands. Openepi is a green software, that stays online 'http://www.openepi.com/v37/Menu/OE_Menu.htm'. You do not need to install it in your computer or laptop. You can visit the webpage, click 'sample size' menu. There are multiple options, each option has a sample case you can check if it is similar to your case. You also can calculate EC50 / IC50 and affinity coefficient of the drug with 'dose-response' feature. I hope you enjoy the online software.
I think you should previously select your primary objective. If you have more than one, I would select the most difficult to demonstrate differences. The next step would be to define the hypothesis (the mimimum difference you may consider statistically significant between two molecules and its deviation). Then you can get different software to determine the sample size from the web.
Greetings,
With regards to the type of statistical test to use, you first should consider the type of measurement of your markers: will you be measuring them on a continuous scale (e.g. concentration) or a discrete scale (e.g. ordinal expression score, dilution, presence/absence of effect)? The ANOVA is suitable only for a Gaussian distributed continuous outcome variable, and you may need to mathematically transform your variables in order to comply with this requirement. If your measurements are on a discrete scale, no transformation will make them suitable for an ANOVA: you should consider the use of a generalized linear model that uses the appropriate link function for the tested distribution (my first choice), or a non-parametric version of the ANOVA, e.g. a Kruskal-Wallis test (a far second choice as the test may produce unreliable results with skewed data and/or unequal variances among treatments).
Second, since you are measuring 3 outcome variables for all 6 drugs, you should consider the probability of type-I error for one of the markers. Therefore, you should consider using a Bonferroni-corrected significance threshold not only for the post-hoc comparisons, but also for the omnibus test. If all your outcome variables are Gaussian, then you may use a MANOVA.
Third, with respect to post-hoc testing, there are multiple options: the most important thing is to prioritize the comparisons on which you are most interested. Is there a reference molecule to which all the others should be compared? is there a scientific basis for performing all pairwise comparisons or just a subset of them?
Finally, with respect to sample size, it depends on the size of differences in effect between molecules. If you have a negative control group (i.e. drug-free vehicle), you should reach statistical significance even with triplicate testing. But if you have a positive control molecule, maybe you should consider a larger number of replicates.
I hope this helps. Good luck with your experiments!
The element of the sample can be anythin such as lab result, humban beings, records..etc. based on the research question or hypothesis, one will determine the the statistical test needed.
To do power analysis to estimate your sample size, you have to write your hypothesis, and based on that you decide what statistical test you will use. It should be one of the inferential statistics. so you need to determine the following: alpha {standard to be .05}, power [standard to be .80], effect size {small, moderate, or large, each test has its own value, you can find these values in the net}. Then download free programs to calculate the sample size such as G. power.
- Badges
- Science topic