Alternative to Kruskal-Wallis

#1
Hi everyone ,
I am currently run a Kruskal Wallis(KW) analysis relating constructs with a demographic info. ( example i like to go to school (construct) with state district ( i have 4 districts ) . The constructs are Likert Scale items with 5 choice of questions. However the construct was a member to one large theme/section ( example : Children and School). There was more constructs contained in the sections ( another 9 of them) Right now , i need to do analysis as in total of one chunk of section against state districts. In other words, rather than looking for significant relation between a construct and districts, i want to look for significant relation between 9 construct ( 1 theme) and districts. Should i consider any other test than KW ? Thank you in advance .
-AZ-
 
#3
Could it be justified to collapse all 9 measurements into 1?

With kind regards

Karabiner
I did try , i sum up all score of 9 constructs with min scores 9 and max 45 . (1) Somethng wrong with the way i key - in the scores, SPSS cant seem to read them (2) i am not particularly sure if its academically legit
 

Karabiner

TS Contributor
#5
(1) Somethng wrong with the way i key - in the scores, SPSS cant seem to read them
I do not know what you mean, sorry.
(2) i am not particularly sure if its academically legit
Well, that is often the case. If the 9 variables are meant to jointly represent the same
phenomenon, than this could be a justification. Maybe you should try to measure
the internal consistency of your scale (Cronbach's alpha), as a start.

With kind regards

Karabiner