OK, I think that I now understand the situation. I hope that Jim Clark, Karl Wuensch, and others are paying attention, because I trust their responses more than my own.
Here's what I understand so far:
You summed data over 9 questions on a 4-pt scale. Your mean, sd, and n were: 24.37, 5.64, and 190. You'd like to compare your results with data from a study done with the same 9 questions on a 5-pt scale. Their mean, sd, and n were: 26.37, 12.68, and 54. Your data are accessible, but you have access only to the summary data from the UAS study.
I think that you would not be able to make a statistical comparison without making many assumptions (most of which would be challenged by the research community). Your biggest problem is that you used a 4-pt scale. Had you kept the same structure as the original scale, I think that a comparison would have been reasonable. To me, the issue goes beyond the simple assignment of numbers to the scale points (which could be open to challenge, even though everyone does so). With 4 points, a respondent cannot choose a mid-point, but is forced to one side or the other. With 5 points, a respondent can choose a neutral response. Thus, I cannot think of a way I could re-scale your data that would allow me to treat the two studies as comparable. (That's what I meant by looking for a "fair" comparison. Transforming to percentages [24.37/36 vs. 26.37/45] would only work if you could assume that your mean wasn't elevated by forcing people off the mid-point through the use of a 4-pt scale. I don't think that's a safe assumption.)
I wouldn't let my inability to compute a statistic prevent me from comparing the two studies, however. For instance, I presume that you didn't collect the data in isolation, nor did the original UAS study. Do you have demographic data (gender, etc.) that you could use to break down your responses? Did the UAS study look at such breakdowns? If they didn't report such information, then I do think that it might be useful to email the author of the UAS study to see if the author would send you the original data set. That way, you could at least determine if the breakdowns produce similar or different patterns in comparing the two data sets.
I hope that you can understand my ramblings. If not, then email me directly and I'll try to clarify.
Hugh
On Sunday, December 22, 2002, at 08:44 PM, 赵玉芳 wrote:
Dear prof. Hugh:----------------------------------------
Thank your.
My English is lack of fluency and I seldom use it write something, just reading. so the word and express I used is not good enough.
The question I asked came from a real study. Someone translated a English scale into Chinese. At first I did not think compare the data. There no correspondence word to translate the scale's five choices. In common sense a 4 choices item is more preferable to the Chinese subject, it is easy to answer.
After I collect the data, someone tell me, the questionnaire came from English, why not compare your data to the data come form UAS. My god! a 4-pt scale compare a 5-pt scale, I never seen anyone do it, and there no original dada come form UAS yet, there only M and SD.
What should I do? I asked the TIPS only find a possible solution, if there no, bad luck.
It's not a homework my teacher assigns to me, just a work of my self.
About the 3 question your think ,I may answer some of them:
no1. I only try to see is there any sig. deference between the UAS and Chinese subjects in the same scale (but one 4-pt one 5-pt).
no2. The comparison may not "fair", although the mean are all summed over 9 questions.
no3. I don't know this question I can answer or not.
======= 2002-11-29 14:11:00 ：=======
I'm sorry, but you will learn best if you do your own work.
Here are some questions I might ask myself in thinking through such a
problem:
1. Why do you want to compare the data? What are you trying to learn
about the data?
2. How do you make your comparison "fair"? What would be the impact of
different sample sizes and different scales (4-pt vs 5-pt)? Were both
means summed over the same 9 questions?
3. What statistic seems most appropriate, given your question and what
you know about population parameters?
Good Luck!
Hugh
On Tuesday, November 26, 2002, at 08:47 PM, 赵玉芳 wrote:
Hugh Foley,！
I forget it, the mean of 24.37 summation over 9 questions,my
sample size is 190,the correspondence sample size is 54.
It is a very important homework for me.I overlook the original
questionnaire is a 5-pt scale,and i used it as a 4-pt scale .Oh now I
dont know how to compare !
Thank you help me!
======= 2002-11-26 09:27:00 =======
Homework problem? How does a mean of 24.37 emerge from a 4-pt scale?
Summation over many questions?
Hugh
On Tuesday, December 17, 2002, at 10:45 PM, 赵玉芳 wrote:
Hello,
I have a problem.I collected some data with Likert 4 point
scale,and there also some data about the same questionnaire but
collected with likert 5 point scale and there have no orignal data
yet,I only can see the Mean and SD . Now how to compare them?
e.g. my data's Mean and SD are 24.37 5.64
the correspondence data's Mean and SD are 26.37 12.68
Is there some one would like to tell me how to compare the data.
Thanks.
[EMAIL PROTECTED]
2002-12-18
---
You are currently subscribed to tips as: [EMAIL PROTECTED]
To unsubscribe send a blank email to [EMAIL PROTECTED]
Hugh J. Foley, Ph.D.
Professor
Department of Psychology
Skidmore College
Saratoga Springs, NY 12866
(518) 580-5308
----------------------------------------
