Non Significant Results Dissertation SPSS: What to Actually Do Next

A non significant result in SPSS means your p-value is above .05, so you failed to reject the null hypothesis. It usually traces back to one of four causes: an underpowered sample, the wrong statistical test, a genuine null effect, or a coding error in your data. None of these, on their own, fail a dissertation.

If you searched for non significant results dissertation SPSS because your output just handed you a p-value above .05, you are in the right place. I have been doing statistical consulting and dissertation guidance for over 12 years now, and this exact panic email lands in my inbox every single week. Let me walk you through what is actually happening, because most of what you will read online stops at “don’t worry” and never tells you why it happened.

What a Non-Significant Result Actually Means (And What It Doesn’t)

A non-significant result simply means your p-value came out above .05, which means you failed to reject the null hypothesis. That is the full technical definition. Nothing more, nothing less.

A non-significant result is a statistical outcome where the observed p-value exceeds the chosen alpha level, usually .05. It means there is not enough evidence in your sample to confidently say the effect you were testing for is real.

Two error types sit behind every hypothesis test, and knowing the difference matters more than students realise. A Type I error means you claimed an effect that was not really there. A Type II error, the one quietly hiding behind most non-significant dissertation results, means a real effect existed but your study missed it.

Here is the part nobody explains clearly enough. Non-significant does not mean “no effect exists.” It means your data did not give you enough evidence to detect one, with the sample and method you used.

Can You Pass a Dissertation With Non-Significant Results?

Yes, and in most universities across the USA and UK, you can pass comfortably. Committees mark you on the rigour of your method, not on whether the universe cooperated with your hypothesis. I have sat through defences where the non-significant result was the stronger, more defensible chapter, simply because the student understood exactly why it happened.

What genuinely gets students in trouble is not the non-significant result itself. It is not being able to explain it when asked. That is what the rest of this article fixes.

And no, it is not too late to fix this. You do not need to re-collect data at this stage, and a supervisor worth having will not think less of you for diagnosing this honestly instead of hiding it.

The 4 Real Reasons Behind Non Significant Results in Your Dissertation (SPSS)

If “dissertation results not significant” describes your SPSS output right now, you are usually looking for one of four explanations, not just one. Most articles covering this topic give you comfort and stop there. Comfort is nice, but it does not get your results chapter written.

In my experience, every non-significant result traces back to one of exactly four causes. Here they are, in the order I check them with my own students.

Reason 1: Your Sample Was Underpowered

This is the most common cause by a wide margin. If your sample size was decided by convenience rather than a proper power calculation, your study may not have had enough statistical muscle to detect a real effect, even if one exists.

Your power calculation needs an expected effect size, usually expressed as Cohen’s d for a two-group comparison. This tells you how many participants you actually needed. Free tools such as G*Power let you run this in under five minutes, using the sample size and effect size you actually have.

A quick way to check: if you never ran a power analysis before data collection, run one now using what you already have. The statssy sample size calculator will tell you in minutes whether your study was ever capable of reaching significance.

If your power comes out below 0.80, that is your answer. And if your sampling method was convenience sampling, it is worth checking whether that choice is still defensible in your limitations section too.

Reason 2: You Ran the Wrong Statistical Test

This one quietly ruins more results chapters than people admit. A Pearson correlation run on ordinal Likert data, or a parametric test used on non-normal data, can mask a real relationship that the correct test would have picked up.

Before you accept your result as final, check the assumptions of the test you used. The mix-up between Pearson and Spearman correlation is the single most common error I see in dissertations, and it changes p-values more often than students expect.

If you are running post-hoc comparisons after an ANOVA, the same logic applies. Choosing between Tukey HSD and Games-Howell depends entirely on whether your group variances are equal, and getting this wrong can quietly bury a real difference.

Reason 3: There Genuinely Is No Effect

Sometimes the honest answer is that your hypothesis was wrong, and that is completely fine. Research is not a prediction contest. A genuinely null result, reported honestly, is a legitimate scientific contribution, and this is exactly what a null results dissertation chapter looks like when it is done properly.

I want to be blunt about something here. A lot of the most-shared guides on this topic, including GradCoach’s widely read piece on non-significant results, do real good by normalising the panic. But they stop exactly where the diagnostic work should start. They never walk you through checking your own power, your test choice, or your data coding.

Comfort without diagnosis gets you through one bad week. It does not get you through your viva. GraphPad’s statistics team has actually documented five distinct technical explanations for a non-significant result. Most dissertation-focused advice online ignores all of them in favour of pure reassurance.

Reason 4: A Coding or Data Error Is Hiding Your Real Result

Nobody writes about this one, which is strange, because it happens constantly. Reverse-scored Likert items that were never recoded, a continuous variable accidentally read as categorical, or missing data handled inconsistently, all of these can erase a real effect without you ever knowing it was there.

Multicollinearity is a sneaky variant of this problem in regression models. Two predictors overlap so much that neither one shows significance alone, even though together they clearly matter. The Variance Inflation Factor (VIF) guide explains exactly how to catch this before you conclude your variables don’t matter.

How to Check Which Cause Applies to You: A Quick SPSS Walkthrough

Here is the exact sequence I use with students, step by step.

1. Open your Sig. column first. Confirm you are reading the right p-value, two-sided unless you specifically pre-registered a one-sided test.
2. Check your assumptions table. Normality and homogeneity of variance violations can push a real effect into non-significant territory. The Shapiro-Wilk normality test guide shows you exactly which SPSS output to look at.
3. Run a post-hoc power check, with caution. It will not prove your study was adequately powered, but a very low number is a strong hint.
4. Review your variable coding. Open your data view, not just the output, and confirm every reverse-scored item was actually recoded before analysis.
5. Re-check your test selection against your data type one more time, with fresh eyes.

Nine times out of ten, this sequence tells you exactly which of the four causes applies to your dissertation.

How to Write Up Non-Significant Findings (APA Style)

Here is how to write up non-significant findings without sounding like your study failed, because it did not.

In your results chapter, state the test, the statistic, and the exact p-value plainly. For example: “An independent samples t-test revealed no statistically significant difference between groups, t(48) = 1.12, p = .267.” No apology, no hedging.

Alongside your p-value, report the confidence interval. It tells your reader the plausible range of the real effect, which matters more than most students realise, and it is something the APA’s own reporting guidelines explicitly recommend.

In your discussion chapter, this is where the real work happens. A discussion that says “this may be due to an underpowered sample, as the post-hoc power analysis indicated power of 0.52” shows far more academic maturity than a vague “further research is needed.”

One phrase to delete permanently from your vocabulary: “approaching significance” or “marginally significant.” A p-value of .049 and .051 are, for all practical purposes, the same number.

Case Study: Defending a Null Result in the Viva

A student I worked with, let’s call her Ritu, came to me three weeks before her viva. Her dissertation was full of non-significant correlations between workplace autonomy and job satisfaction. Her sample size was 52 employees.

She was convinced her research had failed. We ran the numbers together and found her power was sitting at 0.58, well below the 0.80 threshold. Her sample simply was not large enough to catch anything smaller than a fairly large effect.

This is exactly the kind of null results dissertation story nobody shares, which is a shame because it is often the most instructive one. Instead of hiding the low power figure, we built it directly into her discussion chapter alongside a recommendation for replication with a larger sample.

In her viva, when the panel asked why her hypothesis was not supported, she had a specific, confident, numbers-backed answer ready. She passed with minor corrections, and the examiner’s note on her discussion chapter read that her engagement with the limitation was “exemplary.”

What NOT to Do: Fishing, Rerunning, and p-Hacking

Do not rerun the same test with a different subgroup, a different covariate, or a different outlier-exclusion rule just to chase a lower p-value. This practice, known as fishing for significance, inflates your false positive rate and is easy for an examiner to spot once they ask to see your full analysis log.

Do not switch from a two-tailed to a one-tailed test after seeing your results. Decide your test direction before you look at the output, not after.

When to Get Help vs When to Push Through Alone

If you have worked through the four causes above and you are confident in your diagnosis, you can usually write this chapter yourself with the structure given here. If you are still unsure which cause applies, or your power analysis and test assumptions are throwing up conflicting signals, that is exactly the point where a second pair of trained eyes saves weeks.

The quickest way to decide which situation you are in is to read how to know if you need a dissertation expert before you book anything. It walks through the same diagnostic thinking used in this article, applied to your specific chapter.

FAQ

About the author: Written by Siddharth Gupta, 12 years of analytics consulting and dissertation statistics guidance across R, SPSS, Python, and Stata. Connect on LinkedIn