# Why does my research findings contradict my hypothesis result?

From my research findings/results it was clear that lecturers and students use different web 2.0 applications. But my null hypothesis result is contradicting this one of my null hypotheses is 'there is no significant difference between the web 2.0 application commonly used by students and those used by lecturers'.

-
What do you mean by you research/findings and what do you mean by "null hypothesis result" - do you mean it wasn't rejected? How is your research different from your hypothesis test? –  Corone Mar 6 '13 at 8:31
Hi Kiki, welcome to CrossValidated. In order for people to give you good answers they need enough details to understand the issue. Also, CV is not twitter - there's no 140 character limit, so please avoid abbreviations like 'Bt' and 'Pls' - make life as easy as you can on the people trying to understand what you want. –  Glen_b Mar 6 '13 at 8:55

I am not sure I understand your question, but I think I see what is confusing you. The null hypothesis (usually something like "there is no difference", "there is no relationship") is what you want to reject. If your results indicate that there is a difference, then you reject the null (that is the desired outcome of a test, nearly always). If the data do not show a difference, then you fail to reject the null. A failure to reject the null could be due to low power, poor measurement, a bad theory or other things.

-

Because statistics can only be used to reject hypothesis, it cannot be used to "accept" a hypothesis or prove that a certain hypothesis is right. This is due to the limitation that we can only estimate the distribution of an underlying parameter if the null is true (in your case, the proportion of web devices used being equal between the two groups in the population level). We cannot guess the distribution of the same parameter if there really is a difference in the proportions, because no one knows where the exact difference lies.

So, instead of proving what we want to prove, we use statistics to reject the opposite of what we want to prove. If you can reject "there is no difference," then you would conclude in favor of the alternate hypothesis that there is indeed a difference.

-

Just because the results look good, doesn’t mean they are statistically significant. A hypothesis test tries to determine whether, statistically, one can expect the same results from a repeated study. Whether the null hypothesis will be rejected or not, will depend on the estimated variance in your experiment, the size of the difference in proportion as well as the number of samples used.

The null hypothesis states the results obtained are purely random, whereas the alternative hypothesis states that the results are extraordinary. That is, at a 95 percent confidence level ($\alpha=5$ percent), the null hypothesis will only be rejected if results is expected to be ''that good'' or better in just 5 out of 100 cases. That is, the null hypothesis is rejected in favour of the one-tailed alternative hypothesis is if $p(t \ge t_d | H_0)\le \alpha$ where $t_d$ is your test statistic.

This can also be interpreted as, the zero hypothesis will correctly rejected the null hypothesis 95 percent of the times. Incorrectly rejecting the null hypothesis is called a type I error. The probability of a Type I error = $\alpha$.

-
-1 A successful hypothesis test (with p=0.05) actually says: If the null hypothesis were true, your result would only look that good in at most 5 out of 100 cases. –  ziggystar Mar 6 '13 at 9:58
@ziggystar, your result would only look that good in at most 5 out of 100 cases I'd prefer to put it somewhat differently: "your result is expected to be that good or better in just 5 out of 100 cases". –  ttnphns Mar 6 '13 at 10:54