I am doing a study on the Barnum effect and have my results but a bit undecided which test I should I use.
I have tested 40 people to compare if there scientific level of education effects their susceptibility to the Barnum effect
People have three options GCSE, A-level or graduate and then they rate from 1-5 how much they think the personality is about them.
My teacher said most of the group would be using mann-whitney U but that does not work so well for three options, he suggested I could use Anova instead?
Anyone else have some thoughts what I could use?
Whichever route you choose, include your rationale for your methodology in your write-up.
Ok... now if you want help you need to give us more information.. eg sample size.. groups - and number of participants in group.. what is your dependant and independent variables.. and most importantly what is your research question...
I appreciate your effort and that you don't have experience in statistics but if you want to do it right you need to understand the principles of it and you need to correct your use of terminology...
Now, Mann-Whitney U is an alternative to independent t-test... It is a NON-parametric test, e.g used when the assumptions for independent t-test are violated.. mostly the normal distribution assumption, which normally happens in not large enough samples... That said normal distribution is what you certainly want to check, along with missing data and outliers.. For the distribution, better option is to rely on graphic representations such as histograms as often the statical tests are quite sensitive to small deviations of the mean and SD, and thus yield a significant result, which means the assumption is violated... However, even if a test gives significant result but the histogram is clearly normally distributed, I would say it is safe to carry a parametric test e.g. independent t-test.. Again normality can be corrected using some advanced techniques depending on how skewed the data is, you can transform the data set, but that is not for your level at the moment (worth looking up if you want to do research methods)...If you use ANOVA... again you need to check the assumptions and if needed use non-parametric version of the test, and most importantly understand and discuss what a non-parametric results mean for your research...
What test you choose depends mainly on what question you want to answer, without this information I am afraid, I cannot help much...However, looking into your description you have one IV with three levels and one DV:
one - way ANOVA - interval & normal data
Kruskal Wallis - interval and ordinal
if groups are independent or for repeated measures:
One way repeated measures ANOVA - interval and normal
Friedman test - ordinal and interval
repeated measures logistic regression - categorical data
as your IV - levels are categorical you will need to dummy code your data
Pearson's is testing for linear relationship between variables it is a parametric test.. Spearman's is the non- parametric version.. Anyway they do test correlation which means they treat variables the same, thus one is not dependant upon the other, so it will not be appropriate for your question does level of education affect Barnum effect, however it can tell you if they are both correlated and what is the direction of the relationship if one exists.... You will find a lot of researchers using correlation and implying causation , but the truth is this test cannot tell you that. So I will say correlation is not appropriate for you...
Also, just an advice review when to use the words "affect" and "effect" as well as there, they're, and their as you have a problem in that area... Also you cannot prove or disprove a hypothesis especially in Psychology, you can provide evidence in support for your alternative hypothesis or accept the null hypothesis..
Don't want to confuse you any more, but a non significant result does not necessarily mean non-significant findings, here comes the role or the confidence interval, critique, and improvement of your research design...