How to test data if it is normal distribution in spss ?

#1
Hello
Good morning


guys i will need some help truly i have variables in spss most of them scale type, some of them are nominal, and few are ordinal.

My questions how can i get the graph for normal distribution for the whole data ? shall i combine all scales into one variable and combine all nominal and ordinal into one variable or what ?.

i watch couple of videos they go to descriptive > explore > dependent is the scale and independent the nominal , other way they go to graphs > histogram > they choose display normal curve or something like that.

What is the correct way to check my whole data if they are normal distribution or not ?!
 
Last edited:

Karabiner

TS Contributor
#2
You cannot check a whole dataset for distributional
assumptions, only single variables. Or, at least, there
is no need to check the whole dataset. Or why do
you have such thing in mind?

By the way, normality assumptions apply ONLY to
interval scaled variables. Moreover, nearly always it is
absolutely unintersting to assess normality of raw data.
There is no standard statistical procedure which requires
normal or near-normal dirstribution of raw data.

With kind regards

K.
 
#4
Of course it is possible to check if a variable is normally distributed or not. Just do a histogram and overlay it with a "normal curve".


By the way, normality assumptions apply ONLY to
interval scaled variables.
I must say that I have not thought about it, but I thought that there was no relation between if a variable is ordinal or interval and if it was normally distributed. If a variable is continuous does not mean that it is interval. If we have ordinal variable about "liking" or "satisfaction" on a visual analog scale (VAS) then that variable could be approximately normally distributed. Couldn't it?

Isn't it possible to take any continuous skewed variable, take it's median and define that as mu (the expected value in a normal distribution) and define the deviations from mu so that it fits to the normal distribution function. But that would not make the distances to interval scales (as I understand it).

- - -

Maybe it is better if the OP explain about the problem instead the detail of normality.
 
#5
Hi, Thanks for reply
yes i know i can show the histogram and normal curve but for only each separate item will show histagram and table etc.


My data is most of them scales, few nominal , if i want to show the normal shall i put in dependent all (scales) ? and independent all (nominal) or this is wrong since as u sayed there is no releationship between scaes & nominals.


Also one other questions do i really need to choose the independent varaible in explore ?

Can someone show me the steps to do the normality test either in SPSS or other program ?
 

Karabiner

TS Contributor
#6
If we have ordinal variable about "liking" or "satisfaction" on a visual analog scale (VAS) then that variable could be approximately normally distributed.
Well, AFAIK either one assumes that the VAS was successful in
representing liking or satisfaction on an interval scale, so
that a normaliy (etc.) considerations make sense. Or, one
assumes that liking or satsfation are truely ordinal and that
VAS only looks like measuring on an interval scale, but in
reality isn't, in which case normality considerations seem
pointless.

With kind regards

K.