# Thread: Normality plot (no of errors over 2 years)

1. ## Normality plot (no of errors over 2 years)

I am trying to analyse number of errors over a period of 2 years and I run a normality test for my data via minitab.

it shows

Can i conclude that the graphical representation show non-normal data ?
then what should i do ?

Thanks!

2. ## Re: Normality plot (no of errors over 2 years)

If that is a Q-Q plot then your data is not normal.

3. ## Re: Normality plot (no of errors over 2 years)

Month No. of errors occurred
Jan 0
Feb 5
Mar 1
Apr 0
. 1
. 4
.

etc...

that's all. Do you need the full data ?

4. ## Re: Normality plot (no of errors over 2 years)

Any tip from anyone ??

5. ## Re: Normality plot (no of errors over 2 years)

i would like to test for special causes, but if my data is non-normal , is there any way out or I need to gather at least 30 for my sample size ?
i can't get 30 as I only have data from 2010. and the number of errors made is less than 30.

6. ## Re: Normality plot (no of errors over 2 years)

Plot also histogram of your data. You have most of your data =0 which indicates normal dist is not a candidate and perhaps a right-skewed distribution is need like chi-square ..etc. Also check Poisson distribution. Please post a table of your complete data.

7. ## Re: Normality plot (no of errors over 2 years)

This is my dataset :

Month No of Errors
Jan 2010 2
Feb 2010 3
Mar 2010 1
Apr 2010 1
May 2010 1
Jun 2010 4
Jul 2010 1
Aug 2010 1
Sep 2010 0
Oct 2010 2
Nov 2010 3
Dec 2010 1
Jan 2011 2
Feb 2011 2
Mar 2011 1
Apr 2011 0
May 2011 5
Jun 2011 0
Jul 2011 0
Aug 2011 0
Sep 2011 1
Oct 2011 0
Nov 2011 0
Dec 2011 1
Jan 2012 0
Feb 2012 0
Mar 2012 0
Apr 2012 4
May 2012 1
Jun 2012 1
Jul 2012 1

8. ## Re: Normality plot (no of errors over 2 years)

I managed to get 2010 data till July 2012,so i believe the sample size (N=31) is now sufficient.

9. ## Re: Normality plot (no of errors over 2 years)

A histogram of your data indicates right-skewed data:

Normal Q-Q diagram is:

And variation of data form normal:

To test whether data fits a normal distribution or a Poisson distribution:

H0: data is normally distributed :: reject (p<0.05)
H1: data follows a Poisson distribution :: accept (p>0.05)

Data can also follow other distributions. You should check and select the best that fits and describes your data.

Please notice that at p=0.01 significance level your data can be considered marginally normally distributed but not at p=0.05 significance level. However, usually as you collect more data these conditions will change.

10. ## Re: Normality plot (no of errors over 2 years)

Problem with your Q-Q is that you plotted the data points while you should plot the Expected values.

11. ## Re: Normality plot (no of errors over 2 years)

Originally Posted by hazel
This is my dataset :

Month No of Errors
Jan 2010 2
Feb 2010 3
Mar 2010 1
Apr 2010 1
May 2010 1
Jun 2010 4
Jul 2010 1
Aug 2010 1
Sep 2010 0
Oct 2010 2
Nov 2010 3
Dec 2010 1
Jan 2011 2
Feb 2011 2
Mar 2011 1
Apr 2011 0
May 2011 5
Jun 2011 0
Jul 2011 0
Aug 2011 0
Sep 2011 1
Oct 2011 0
Nov 2011 0
Dec 2011 1
Jan 2012 0
Feb 2012 0
Mar 2012 0
Apr 2012 4
May 2012 1
Jun 2012 1
Jul 2012 1
1. It cannot be normal as it is discrete.

2. Why would you even expect it to be "normal", you should be expecting it to have a Poisson distribution which would only be approximately normal when the mean number per period is large (which it is not).

.

12. ## Re: Normality plot (no of errors over 2 years)

Originally Posted by MaxJasper
Problem with your Q-Q is that you plotted the data points while you should plot the Expected values.
Sorry i don't understand what do you mean. Could you please explain in detail ? Thanks

13. ## Re: Normality plot (no of errors over 2 years)

This is what I have got from minitab: Graph > Probability Plot.

14. ## Re: Normality plot (no of errors over 2 years)

With Reference to
Identifying the Distribution of Your Data - Minitab

I do a check and I got this :

P<0.005 so it is not normally distributed. Where do i go from here ?

15. ## Re: Normality plot (no of errors over 2 years)

No one ?

Page 1 of 2 12 Last