Hi Karen,
First of all, I do hope to see one day that we have a consensus towards how we all do Statistics.
This is how I use the skewness and Standard error.
- you will need to use Analyze-->Descriptives Statistics-->Explore first.
- Check the table to find the skewness and standard error values.
- Use the formula (skewness-1.96*(standard error),skewness+1.96*(standard error)). This is to find the 95% Confidence Interval for skewness. 1.96 is the critical value of Z for 95%.
With this boundary,(skewness-1.96*(standard error),skewness+1.96*(standard error)), you will need to check if 0 is contained in it. If it does, it just means that skewness is approximately 0. if it does not means that skewness is not approximately 0.
You will need to repeated the same step for Kurtosis as well to check if it is approximately 0.
Now remember the checklist that I have provided?
-
§Bell shaped curve?
-
§Is the distribution symmetrical? (Histogram)
-
§Check if MEAN ≈ MEDIAN ≈ MODE (can be check through frequencies in PASW Statistics)
-
§5% Trimmed mean similar to mean?
-
§Check if Skewness ≈ 0
-
§Check if Kurtosis ≈ 0
-
§Check for Outliers (Explore - Box-plots)
If any of these fails, the scale variables is not normal. A tip is not to calculate the skewness and kurtosis first as it takes time to do the calculation so do check the rest first.
Regards
Dorraj
Date: Sun, 6 Dec 2009 10:32:21 +1100
From: [hidden email]
Subject: Q-Q Vs P-P plots
To: [hidden email]
Thank you to everyone who responded to my query concerning the use of Q-Q Vs P-P plots.
Obviously there is no consensus on this issue, but it is statistics after all!!
I am leaning towards the Q-Q plots . However, what I did not mention in my original email is that I never intended to rely on these plots alone. The shape of the distribution (skewness and Kurtosis) and the identification of outliers are/were my primarily concern. As to what qualifies as a departure from normality with respect to skewness and kurtosis I'm going with the actual value of skewness and kurtosis rather than the ratio of each with their SE. Actually while I'm here, what values of skewness and kurtosis are considered to indicate non-normality?
Karen
Windows 7: Find the right PC for you. Learn more.
Locked
