# Normality is not assumed ( kolmogorov-smirnov sig level= 0.000)

#### lindsey chan

##### New Member
Currently, I'm doing a research using Beck Depression Inventory on 340 respondents from my university. Before I apply any parametric tests, I tested the normality first, however the normality is not assumed. So, I wondering is the data really not normal or the way I analyze it have problem ?

My way :
I put the "Beck Depression Inverntory (BDI) level" into the dependent variable box under the EXPLORE menu without recoding any of my other variables .

Variables used :
Gender (Male or Female)
Year of study ( Year 1, 2,3 ,4 )
Races ( malay, chinese, indian, others)
Faculty (FSSA, FKSW)
Courses ( I have 12 diff courses in total)
CGPA (0.00-1.99 , 2.00-2.49, 2.50-2.99, 3.00-3.49, 3.50-4.00)
Residence (On campus, off-campus, with parents )
Monthly allowance
Frequency of getting counseling services ( i used 5 point likert scale)
frequency of attend bonding activities ( i used 5 point likert scale )
frequency of attend motivational talks ( i used 5 point likert scale)
academic perfomance rating ( i used 9 point likert scale)
life satisfaction rating ( i used 9 point likert scale)
Beck depression level ( it has 6 different level of severity)

#### Karabiner

##### TS Contributor
Before I apply any parametric tests, I tested the normality first,
Do you mean normality of the dependent variable(s)? Mind that normality of the dependent variable
is not needed for parametric tests. It is sometimes assumed that the errors from the parametric
model (the residuals, e.g. from a linear regression or from an ANOVA) are normally distributed
in the population, but even that is important only in case of small samples (n < 30 or so).

however the normality is not assumed.
Well, of course not. It is a clinical scale. Why should we assume that a condtion like depression is
normally distributed?

Besides, you have quite a large sample size, so even small deviations from normality
in the population would cause a "significant" test result.

Variables used :
Gender (Male or Female)
Year of study ( Year 1, 2,3 ,4 )
Races ( malay, chinese, indian, others)
Faculty (FSSA, FKSW)
Courses ( I have 12 diff courses in total)
CGPA (0.00-1.99 , 2.00-2.49, 2.50-2.99, 3.00-3.49, 3.50-4.00)
Residence (On campus, off-campus, with parents )
Monthly allowance
Frequency of getting counseling services ( i used 5 point likert scale)
frequency of attend bonding activities ( i used 5 point likert scale )
frequency of attend motivational talks ( i used 5 point likert scale)
academic perfomance rating ( i used 9 point likert scale)
life satisfaction rating ( i used 9 point likert scale)
Beck depression level ( it has 6 different level of severity)
Well, not one of these variables is interval scaled. Therefore it is meaningless to ask for their normality.
For example, one could maybe ask for the normality of the BDI raw score (although I suppose it is not
needed for anything here, see above), but one can NOT ask for the normality of "6 levels of severity",
since this is an ordinal scaled variable. Or consider a binary variable such as gender. How could
"gender male/female" be normally or nonnormally distributed, it is just a binary variable.

With kind regards

Karabiner

P.S.
Frequency of getting counseling services ( i used 5 point likert scale)
Two remarks: a) What you probably mean here is Likert-type item, no Likert scale. There is no such thing as "5 point Likert scale". The 5 points refer to the response scale of a single Likert-type item. Likert scales consist of several Likert-type items, which are summed up.
b) this item can probably not be called a Likert item. Likert-type items measure degree of agreement, not frequencies.

Last edited: