How can I get normally distributed data? I have done Repeated Measure Two way ANOVA for my ample of data using SIGMAPLOT. My analysis has two factors, for example: 1st factor is the hydrogen peroxide concentration and the 2nd factor is about time series (20 minutes with every 1 minute interval I took the absorbance reading).
What happened was my normality tended to fail and even sometimes my variance test failed. Then this SIGMAPLOT can suggest me to proceed with the Holm-Sidaak test. I have never used that or came across in any analysis.
So I need advice on how from the beginning I can collect (how many or replicates) data to make my data are normally distributed when I run a normality test where I can use Tukey or Post hoc test which is commonly used. I would like to run a new set of experiments. How can I proceed?
 A: You can't make your data normally distributed by collecting more of it. 
The rest of your question is very unclear. You don't say what your dependent variable is, what data you have, what your hypothesis is, or anything else, so it's impossible to offer more guidance. But, for your main question, the answer is "you don't". 
A: Two general misconceptions in your approach. First: Data for linear models do not have to be normally distibuted. Residuals preferably should, but with large numbers ANOVA and ANCOVA get quite resistant to deviations.
Second: No variables in real life are perfectly normally distributed and normality tests will therefore fail with all real data, given a large enough sample. Data can be normal enough for any practical purpose and still show significance in normality tests. 
Collecting more data does not usually make them more normally distributed but it makes t-tests, ANCOVAs and many more approaches more robust towards non-normally distributed data and residuals.
A: Basically for my experiment,i will go and collect my raw plant sample from sea for 3 times.I explain what i will do for my first time collection because the other two times it will be same way of analysis (i consider 3 times as a replicates as well).
So for my time collection:
I will process my sample and start from crude enzyme for the analysis:
1) i will do spectrophotometric (colorimetric) analysis in 96 well plate (its just a plate with many wells) .analysis usually with blank, control, 4 replicates of sample (altogether 6 wells for every row)
1.1) firstly, 
a) i will have 3 different composition (A,B,C) and do this colorimetric analysis by time course for 20 minutes. Once i get the absorbance i will calculate and plot the result for this 3 different composition (A,B,C) separately, where x-axis will be time (20 minutes) and y-axis as the activtiy. The 20 minutes reaction i will get reading for every 1 minutes interval. So i will include all the starts from 0 minutes to 20 minutes reading in this comparison.Then i want to compare which 3 (A,B,C) will give a good result.
So, from my understanding if 3 samlpe and by tiime course comparison we will be using two way repeated measure ANOVA. But when i tried to analyse using sigmaplot it falied normality test but passed for variance test and analyzed using "Hold-Sidak method" which give me result satisfying for me but when comparnig these 3 (A,B,C) by time course i didn't see any changes from the graph but this statistical analysis shows significantly different(probably minute changes). but this is not acceptable for the reviewer or my supervisor.
Thus, to get a right analysis (normality test passed) what are the steps i should correct? of number of sample? besides that in sigmaplot i dont know how to create descriptive analysis, as just able to get the variance result and pairwise comparison result
b) Next analysis will be same as before, x-axis is time course for 20 minutes and y-axis will be activity. The analysis is with fixed concentration of 3 different substrates (mixed in one assay).This will be one line graph (presumably linear line or linear and then plateu formed).I'm not sure do i need to repeat the analysis 3 times? because when doing the analysis i have 4 wells as replicates? Do i need to do any statistical analysis here, as im not sure.
