Read. As much as possible (articles and academic books). Your understanding of things and your writing level will improve based on how much you read. You will find inspiration, ideas, solutions and questions. Imagine you need n ideas to write a good thesis and to retrieve them, you need to read at least n articles.

And whenever you have the chance, discuss with colleagues (if they are more prepared than you, better). Having feedback is a treasure for any research.

• Choose a chairman with whom you can get along (ask other students).
• Do a very extensive literature survey, not just the last ten years (but they are especially important).
• When you find an especially interesting and pertinent paper check its bibliography to find similar previous papers. Maybe even go back from there.
• Decide on how you are going to test any data before you collect it. Stick to simple tests. The t test is the best for most simple questions and try to keep your questions simple anyway. For more complicated stuff seriously consider nonparametric tests. Try to avoid the ANOVA especially for more than six replications or blocks. Nonparametric substitutes for the ANOVA avoid most of the pitfalls of the ANOVA, but they also work best for six or less treatments.
• I recommend seeking an 80% power for all tests, and an alpha (probability of a false rejection) of 5% for ANOVA-like tests, and an alpha of 1% for small tests. That 1% also applies as a minimum for multiple comparisons (small tests following the more complex tests).
• Learn enough suitable statistics to be able to estimate sample size. Estimate the sample size before starting to collect the data. And do not examine the data you are going to test before using it to test. Variances etc. can only be drawn from data obtained in exactly the same way as the data you expect to collect.
• Sample size estimation is tricky and tends to give an answer that is larger than you would expect. Samples of size 30 are really too small to detect any difference but huge differences.
• You almost never need to test for normality for a t test with 80 or more samples or pairs. (There are a few instances in which the t test will not work, such as Likert stales.) The nonparametric alternatives to t are true small sample tests and are likely to suffer extra loss in power above sample sizes of, say 80, or more. But small samples are often too weak to detect anything.
• If you must estimate sample size for the t test or whatever parametric test, you may have to search old records for data like you intend to collect (you can not examine the data and then test it), or you might need a dress rehearsal (a small pilot test to estimate the variance) the first year of your masters study. Then you will have to decide how small a difference is a reasonable minimum difference (how small you need to detect). Smaller costs more in time, resources, and money. It is impossible to test for infatesimally small. That is you may have to consider sample size almost immediately even if you intend to do your main data collection next year.

All this does seem difficult, but they do not call those degrees 'masters' degrees for nothing. Best wishes.

Dear colleagues @Jose Alberto Lopez Aleman and @Lewis Maccarter explained very well the procedure to get prepared to start a master thesis. I am also in agreement with them and insist on a good literature search on selected topic especially recent published papers and using the references of that papers for getting further knowledge about the topic. Then thinking about the data to design an appropriate experimental design. When you set these important steps as well, then you can go directly towards the results you want to achieve. Good luck.

Thanks a lot for your invaluable suggestion, Jose Alberto Lopez-Aleman. Noting the point - 'Having feedback is a treasure for any research'.

Great to hear the details from you Lewis Maccarter. Marking 'All this does seem difficult, but they do not call those degrees 'masters' degrees for nothing'.

Thanks, Elaheh Daghighi for your suggestion, well noted.

Ensure you create time for each an every part of the thesis work. You should continue doing literature review as new knowledge is added every hour.