2

I'm writing my PhD thesis. I used a number of statistical tests to test my hypothesis. Only one test was able to answer my question while the rest (3 tests) couldn't.

If I report the results of the tests which failed to answer my research question, it becomes a large number of tables and graphs. My query is whether I should report all the tables and graphs of those statistical tests which could not answer my research question or should I report only the ones which could answer my research question.

  • 8
    Your question seems to be specific to your advisor. Why not ask them? Also, you seem to be cherry picking results, which can be dangerous and is bad form. – Richard Erickson Dec 6 '19 at 19:39
  • 6
    As @RichardErickson says, cherry-picking statistical results is not only poor form, but easily/often results in false conclusions. I recall a student who said that it took a long time to statistically verify his result, because the first 20 or so statistical tests indicated no correlation. But he finally found a test which did... Ok, so he actually proved the opposite of what he thought... but did not realize it. – paul garrett Dec 6 '19 at 20:03
  • 7
    It's worrying that someone is in the writing stage and doesn't understand the answer to this question. – Azor Ahai -him- Dec 6 '19 at 22:23
  • 2
    Hmmm. Do you mean that only one of the tests gave you the answer that you wanted to get? If that is the case, then your thinking is on dangerous ground. – Buffy Dec 7 '19 at 1:44
  • 1
    You need to discuss all the tests in the body of the thesis. You may be able to put some of the tables and graphs in an appendix to avoid breaking the flow of the thesis, but unless you have a good reason you can explain to do otherwise you should treat the tests the same. – Patricia Shanahan Dec 7 '19 at 5:00
9

The comments are correct when they say that it is absolutely wrong to try statistical tests until you find one that you like and then report just that one.

That said, there's a little ambiguity in your phrasing

Only one test was able to answer my question while the rest (3 tests) couldn't.

If they couldn't answer for some reason you can explain that's not simply "I didn't like the result" then you could discuss those tests, explain why they were unsatisfactory, and perhaps provide the tables and graphs in some kind of supplement.

Whatever you do, you cannot just ignore them. You tried them and you must say so.

For all of this - talk to your advisor.

| improve this answer | |
0

If I understand the situation, there are a few options I would suggest:

  1. Inconclusive tests/results should be briefly mentioned, but maybe not shown. If they provide no information, then they're basically clutter text. Here I'm including info that provides no insight on the result you want to show, that is, they neither support nor refute your claim. They are just useless info that someone might wonder whether you did try to obtain, hence why the brief mention.
  2. Controversial test/results are a different matter. Here I'm considering info that disputes, refutes or weakens your thesis. Those are better shown and explained. They provide concessions to your claim that you'll rather make yourself than to let anyone else point out and jump to the conclusion that your whole thesis is wrong.
  3. Redundant test/results would be those that support your claim, but if you've already found other means to prove your point and pass the message through, those can be placed in an appendix if they provide marginally new info, but mostly adding redundant information implies just writing redundant text, it is unnecessary and should likely be avoided.

Also remember the value of conciseness, most people who might read your thesis, which include researchers that could cite your work or the committee for your defense, will likely check the number of pages before reading it, and often complain if they find a big number. If necessary to write an extensive work, then so be it, but if possible, prefer to be more concise, avoiding cluttering and redundancy.

| improve this answer | |
  • Thanks everybody for the enlightenment – Siraj Mazumder Feb 19 at 12:52

Not the answer you're looking for? Browse other questions tagged or ask your own question.