# Thread: Pearson or Spearman - a slightly violated assumption of normally distribution? SPSS

1. ## Pearson or Spearman - a slightly violated assumption of normally distribution? SPSS

Hi!

I am doing a study as part of Research Methods 3 for my Psych degree and have two questionnaires: one is on attitudes to homosexuality, the other on religiosity. I have just got all the data in and am attempting to analyse it, however, I am hitting a bit of a brick wall as to using Pearson's r or Spearman's rho.

I do not have any outliers and there is a linear relationship between the two variables. The only problem is with assessing how normally distributed my data is.

My skewness and kurtosis values are as follows:
Homosexuality
• Skewness: -.357
• Kurtosis: -1.130

Religiosity
• Skewness: .602
• Kurtosis: -.861.

They are all within the +1/-1 bounds bar the kurtosis on the first variable. My stats/SPSS guide states that for Pearson's r to be used the data needs to be 'approximately normally distributed' which it could be argued mine is, but equally that it isn't.

So, I conducted a Shapiro-Wilk which gave me back this:

Shapiro-Wilk
Statistic df Sig.
Homosexuality .927 40 .013
Religiosity .912 40 .004

So there's further evidence for it being normally distributed, but my tutors have told us they would generally prefer us to go by skewness/kurtosis values.

Overall, I'm a bit confused and lost as to what to do. My tutor is a part timer and won't check her emails for a few days yet, so I would really appreciate anyone that can help me!

Becky

2. ## Re: Pearson or Spearman - a slightly violated assumption of normally distribution? SP

how were attitudes towards homosexuality and religiosity measured? i'm willing to bet it's some sort of likert-type scale but it would be nice to know the number of category options before suggesting anything...

3. ## Re: Pearson or Spearman - a slightly violated assumption of normally distribution? SP

1. As spunky is likely getting at, if your data are ordinal (i.e. Likert) you probably shouldn't be using a Pearson test no matter what your normality tests are telling you. You know, a priori and quite definitively, that your data are not normal because they are not continuous.

2. Technically, you should never do a normality test on your data and then choose to do a parametric or non-parametric test on the same data on the bais of that test result, anyway. Either decide in advance which test to use or do a normality test on an exploratory subset of data that is not used for the real test.

3. If the data is Likert you will have to deal with ties. Be sure you are using software for the Spearman test that adaquately deals with ties.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts