[讨论]Scale Development Questions

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

I have 14 questions all touching on the concept of "social consequences".

Cronbach's alpha for all 14 was low, .41.Rotated factor analysis showed 5 factors,these are listed below with the number of variables that composed each. Factor loadings for all variables on their respective component were high, all above .70. friends 4 family 3 belief 3 disorder 2 comm 2

I want to make these into 5 scales/indices. I would like to analyze them as DVs seperately and also combined. I have been taking the standardized z-scores for all the values for each variable and averaging them for each factor, for example:

zfriends = (z1 + z2 + z3 + z4)/4 and so on...

QUESTION:

If I wanted to use all 5 scales back under the heading of "social control," I can just add the average standardized scores (zfriends + zfamily . . .)? there will be no trouble because of the different numbers of variables in each factor?

I am sure that the SAGE book on scale development would have a section on this, but all three copies are checked out!

thanks Matt

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Development questions question Develop tions 讨论 Development questions scale

Matthew,
Once you consider Art's questions you may still need to consider concerns regarding reliability.  If you have poor reliability (Cronbach's alpha is one aspect of reliability) then you cannot have validity.  Thus you cannot be sure what you find is appropriate.  Based on my experience it is relatively difficult to get a really poor Cronbach's alpha if the items are reasonable.  In your case the poor alpha may result from having a small number of items in each subscale (generally having more items in a subscale can increase your alpha), but may also result from poor items.  You may want to examine item correlations to get a sense of the relationships among your data.

You may also want to check out the this book:
Nunnally JC, Bernstein IH. Psychometeric theory. New York: McGraw-Hill, 2005.
and this article:
Clark LA,.Watson D. Constructing validity: Basic issues in objective scale development. Psychological Assessment 1995;7:309-19.


Ian Kudel

Is a very high number of scales to derive from 14 items. What did you use for a stopping rule? Did you try a parallel analysis using the same number of items and cases? did you try rotating fewer factors? Did all of the items have the same response scale? In the reliability analysis did you make sure that items that needed to be were reflected? Were some correlations negative? Art Art@DrKendall.org Social Research Consultants Inside the Washington, DC beltway.