1. ## tests of normality

Hi.

How can I test normality of data? Which is the best tests?

2. ## Re: tests of normality

It depends on why / for what purpose you want to test,
and how large your sample size is. Usually a Q-Q plot is
better than a test.

With kind regards

K.

3. ## Re: tests of normality

Hi.

I have a small group of data. My data is some like this (7,5,0,4,3,2). In this case the best os a QQ-plot?

4. ## Re: tests of normality

I would start with a Q-Q-plot. There also is a very broad rule of normality kicking in with continous data around sample size 30 or greater. But I would be extremely cautious using just that adage alone as a definitive definition.

5. ## Re: tests of normality

With n=6 I think that every test is useless and senseless
due to lack of power. At best, the data are so extremly
different from normality that a test (or the inspection of the
Q-Q plot) rejects the assumption of normality; but a non-
significant test, on the other hand, would permit no inference
about the underlying distribution. One would then assume
that there simply weren't enough data to reject the null.

Again, why do you want to know whether the data are
sampled from a normally distributed population or not?

With kind regards

k.

6. ## Re: tests of normality

My data is like this but with a lot of rows.

Code:
``````  1  1  1  1  1  1
1  1  1  1  1  1
1  1  1  0  0  1
1  0  1  0  0  0
1  1  1  1  1  1
3  4  2  3  3  2``````
But I can have some row with big numbers. And I need to test the normality for each row.

I don´t understand what you say about 30 or greater...we can't use with this?

I want to test the normality for apply some statistical tests after...I know that the big problem is the number of data. But I have other solution to do this?

7. ## Re: tests of normality

Why do you want to test for normality? Why is having normality important for you? Those look like counts so maybe you should just tell us what your overall goal is so we can help you find the best method to do what you actually want to do (as it's VERY rare that a person's interest is just whether or not something is normally distributed...)

8. ## Re: tests of normality

Re: tests of normality
I need to test the normality because I need to apply some tests that I need a a distribution aproximately normal...I need to know if I can aplly...

Other question that I have now is: if I have equal data like

Code:
``2,2,2,2,2,2``
which conclusion about normality I can have? This is aproximately normal?

Thank you

9. ## Re: tests of normality

Looks like Uniform Distribution.

People keep asking you about the tests you will use and objectives because many times the assumption of normality is not on the raw data but on their residuals. They are just trying to help.

Also, all variables have only six observations?

10. ## Re: tests of normality

Yes, all data has only 6 observations.

I'm trying to do a test of normality because I need to apply some other tests (like t-test) and I need the normality of my data... And what I think is: if I have a normal data I apply a t-test, or if not, I apply other test.
Isn't ok?

11. ## Re: tests of normality

You won't be able to reasonably test if data is normally distributed using six observations. Judging from the fact that you have count data though I'll say that your data most definitely isn't normally distributed. Whether that actually matters is a different question though. Can you do us a favor and describe what your data actually is and what you're hoping to test. Don't try to be statistical at all - just tell us what your data actually is and what you want to learn from it.

12. ## Re: tests of normality

I don't know exactly what is specifically the data. Only I know that is a biological data and this frequencies is the number of time that I have some charateristic. I need to know if it's the same distribution in all rows, or if I have some big different cases (I try to explain the better that I can...). But one of my problems is really the quantity of data that I have...

13. ## Re: tests of normality

Are you dealing with RNA-Seq data? If you don't understand the data then why are you the one analyzing it?

14. ## Re: tests of normality

I don't know if it is a RNA-seq. I only know the objective: to detect bif differences in some rows.
But it isn't correct the idea? Test normality and if it's normal apply tests for normal data?

15. ## Re: tests of normality

You say differences in some rows. Does that mean the first three observations are part of treatment 1 and the last 3 are part of treatment 2? Or something similar to that?