Dear Hossein, Replication or repetition do not change the experimental variability. To repeat an experiment, under the same conditions, allows you to (a) estimate the variability of the results (how close to each other they are) and (b) to increase the accuracy of the estimate (assuming that no bias – systematic error – is present).

As a rule of thumb, designs include the repetition (replicate and repetition meaning depend on the scientific field and context) of, at least, one experimental combination. Quite often a center point (in triplicate or more) is repeated. These repetitions allows the estimation of the experimental variability and as such to make inferences about the significance of the effect of the factors under study by comparing them to the experimental variability (noise). However you don’t need to perform those repetitions if you have already a prior and reliable estimate of the variability. Additionally, these repetitions will allow (in certain designs) the assessment of the suitability/validity of the fitted model (opposite to lack-of-fit) . This is done by comparing how well the model predicts the average of the repeated center points. These are the 2 reasons for the repetition of one experiment. Replication (repetition of the complete design) is not common practice unless you need to assess the impact of some uncontrolled factor and/or to confirm the outcome of your analysis (among other less common reasons).

If you need to replicate (complete design) twice or more, in order to estimate a factor(s) is because either the experimental variability is large or the effect is small (or both). Whichever the reason is, replication does not solves your problem! The large variability will still be there until you find out why and fix it. If the factor effect is small you will, most probably, not be concerned with it and focus your attention on the “big” ones.

Hope it helps. Regards Luis

Because is strictly necesary to know / to establish the experimental error.

Inclusive exist a statistical tool named the power of the test that its purpose is to calculate the number of replication, but in this sense you need to know the standar deviation.

Dear Hossein, Replication or repetition do not change the experimental variability. To repeat an experiment, under the same conditions, allows you to (a) estimate the variability of the results (how close to each other they are) and (b) to increase the accuracy of the estimate (assuming that no bias – systematic error – is present).

As a rule of thumb, designs include the repetition (replicate and repetition meaning depend on the scientific field and context) of, at least, one experimental combination. Quite often a center point (in triplicate or more) is repeated. These repetitions allows the estimation of the experimental variability and as such to make inferences about the significance of the effect of the factors under study by comparing them to the experimental variability (noise). However you don’t need to perform those repetitions if you have already a prior and reliable estimate of the variability. Additionally, these repetitions will allow (in certain designs) the assessment of the suitability/validity of the fitted model (opposite to lack-of-fit) . This is done by comparing how well the model predicts the average of the repeated center points. These are the 2 reasons for the repetition of one experiment. Replication (repetition of the complete design) is not common practice unless you need to assess the impact of some uncontrolled factor and/or to confirm the outcome of your analysis (among other less common reasons).

If you need to replicate (complete design) twice or more, in order to estimate a factor(s) is because either the experimental variability is large or the effect is small (or both). Whichever the reason is, replication does not solves your problem! The large variability will still be there until you find out why and fix it. If the factor effect is small you will, most probably, not be concerned with it and focus your attention on the “big” ones.

Hope it helps. Regards Luis

Replication helps in obtaining a better estimate of experimental error.

Hossein,

The unfortunate side effect of replication and repeats is that your experiment run quantity begins to explode in large variable designs. I would recommend performing a factor screening experiment of at least resolution IV to narrow down the important factors. Hopefully you'll be left with a few significant factors where upon performing replicates/repeats for optimization becomes more experimentally/financially efficient. 

Usually, in repeated measures (ie. replications) we observe same subject. For instance, observe 10 subjects 3 three times. In this fashion, we measure the effect between subjects and within subject differently at each case and repetitions. Additionally you can not measure between and within subject variations in 30X1 model but you can obtain some descriptives.

Replication always helps in verifying the results and better estimating the experimental variability and the effects of the experiment's factors.

to explain that using a simple fact: if you understand the reason of sampling units from a polpulation that is exactly the same principle of replication in experiments

Thanks Prof. Luis F. Gouveia,

The explanation is very informative.

Dear Nasser Alhajj , I'm happy you found my contribution useful. That's the point of these forums where we share our knowledge and/or experience with the sole purpose of helping others in need and also to get help when we need it. Regards, Luis