not applicable ANOVA. and no statistical test is advisable, the problem is the two replicas, the p value is sensitive to sample size. For many calculations between alpha and beta (significance and power) should be the minimum size of at least 4 for each of the treatments.

Two Way ANOVA or Repeated Measures ANOVA followed by Dunnett's test or student 't' test............... Hope this should work. Depends upon how many variables you are analyzing at one time.

Gud Luck

Thank you very much, Dr Alok Nahata

It's very convenient to use Dunnett's test, especially when analyzing many variables. However, the result is little bit different comparing to independent sample t-test for some variables/groups. Do you have some ideas explaining this problem?

The ANOVA will tell you whether there is a significant difference of means between the groups, but not which of the groups that differ from each other. If the ANOVA results in a p-value below the threshold value (e.g.

not applicable ANOVA. and no statistical test is advisable, the problem is the two replicas, the p value is sensitive to sample size. For many calculations between alpha and beta (significance and power) should be the minimum size of at least 4 for each of the treatments.

U can use two way ANOVA for two variale sample

I agree with Carlos, ANOVA is not applicable in this situation. But a simple t-test is better held as the sample size is very small.

regards

should consider that any comparison based on continuous variables, as the t-test or ANOVA or other parametric post hoc test, consider the value of the variance. treatment with having two replicates per treatment, and do not forget the outliers, in 2 data is impossible standardize information.

To clarify things, with "replication" to you mean "repeated measures", that is, each person in the control is measured at three different time points?

repetitions or replications is related to the number of experimental units per treatment (according to blocks or other considerations), for whatever type of experimental design used.

As more variables incorporated in the design, the number of repetitions will are increased significantly.

in the case of the repeated measurement is a kind of experimental design (known as related samples) is the case where repetition is measured at different times. The other types of design consider comparisons between experimental units subjected to different conditions (treatment)

We set up the experiment to test effect of some compounds (with different concentrations) on cell culture. We repeated the experiment 2-3 times (of course at different times). Generally, in biology, experiment with 3 replications for each treatment is accepted. However, for some experiments we just can repeat only 2 times. So I wonder if I can do ANOVA, although I can use SPSS to analyze but the statistical result might not be significant.

Two way ANOVA may be used

How many independent variables?,, i see alone treatments (independent variable), exist another variable

For each compound, we consider as 1 experiment, like we set 0, 10, 20, 40, 80 uM and treat the cell culture, and repeat the experiment 2 or 3 times. Then we want to compare the treatment means of those concentrations with the control (0). For some compounds we repeat experiment 3 times, so I think there is no problem to use one-way ANOVA; but the experiment with 2 replications may not give significant result if using ANOVA.

if you look for composite, then ONE WAY, but very possibility that the response is not statistically significant. analyzing considered test power. (1-beta)

By analyzing several compounds simultaneously with different doses, then two way anova (anova in blocks) in a randomized block design

but the problem are assumptions for use it

You can proceed with your regular ANOVA analysis but please be very careful with the results. If you don´t find significant effects of the treatments versus the control, or simmilar effects of two treatments, you could combine reps and increase the power of your analysis. The same is true is one treatment is simmilar to control. It will depend on the variables analyzed too, of course. You did not mention anything about it. For example your can reach to two treatments with 4 reps or 1 treatment with two reps versus a control with 9 reps...

I would definitely ANOVA, not t-test. And, I would check normality first.

If you're using SAS for analysis, LSMEAN would be a good one to start with for telling difference among treatment.

I will go for ANOVA. However, just like Juan mentioned, you may have to combine the reps to increase he power of the statistical test.

Ok. Actually we have 5 treatments (including control). However, we don't seem to agree on number of replications. Choice one: ignore 3rd replication for the one which has got 3 replications. Choice two: you can take average of the two to get a third for those having 2 replications. I would prefer choice one though.

Run one way ANOVA and examine the results. If you don't get significant differences, doesn't matter. You can run Fisher's Protected LSD test for comparing means. And if you get significant treatment differences, include a contrast like 'Control versus treated' to have an idea of the differences.

Hope it helps.

you can used ANOVA as randomized complete design RCB or randomized complete block design RCBD depend on the source of variance in your experiments 

MS Excel can perform two-way anova without replica; 

Nice discussion