Re: [問題] 關於李克特尺度

看板Statistics作者 (小李)時間16年前 (2009/08/18 16:08), 編輯推噓1(1推 0噓 28→)
留言29則, 6人參與, 7年前最新討論串4/4 (看更多)
請教先進有關李克特尺度的問題,最近有一篇期刊的評審問了我一個問題 我在文章裡資料分析段寫道“Before statistic analysis, tests were completed for normality, homoscedasticity, linearity and independence of error terms (Cohen& Cohen, 1983). These assumptions were examined and were found to hold true.” 評審問:I don’t understand the meaning of this explanation. The variables are 5-point Likert scale. Why do they need to be normal, homoscedastic, and linear? Even weirder thing is that these assumptions were found to hold true. Do the authors mean the error terms of the multiple regression model? If so, it shouldn’t come as the first paragraph of the discussion part (even before the factor analysis part). 請問Likert scale的問項在進行迴歸或因素分析之前是否要做這些檢定呢?因為Hair教科 書裡很多分析之前都要先分析是否符合假設,但是沒針對尺度特別說明一定要或一定不要 做這些分析 這讓我蠻困擾的,不知道我有沒有會錯評審的意思?評審是指Likert scale的尺度特性本 來就不必做這些測試嗎? -- ※ 發信站: 批踢踢實業坊(ptt.cc) ◆ From: 122.127.176.224 ※ apson:轉錄至看板 PhD 08/18 17:52
→
08/18 20:47, , 1F
The 5-point Likert scale apparently has a discrete
08/18 20:47, 1F
→
08/18 20:49, , 2F
distribution, but you performed a test to check
08/18 20:49, 2F
→
08/18 20:51, , 3F
whether it has a normal distirbution, which is used
08/18 20:51, 3F
→
08/18 20:51, , 4F
to model the distribution of continuous variables.
08/18 20:51, 4F
→
08/18 22:21, , 5F
statistic analysis? or statistical analysis?
08/18 22:21, 5F
→
08/18 22:22, , 6F
before statistic analysis?原來你那些test都不是統計分析
08/18 22:22, 6F
→
08/18 22:23, , 7F
其實我一直覺得 台灣那些作問卷跟評量的 都像你一樣亂搞
08/18 22:23, 7F
推
08/18 22:31, , 8F
樓上第三句話有點太重,並沒有"都"好嗎?
08/18 22:31, 8F
→
08/18 22:56, , 9F
沒有見到review全文也不知道原作的領域為何
08/18 22:56, 9F
→
08/18 22:57, , 10F
但推測reviewer的意思是Likert scale有其本來的限制
08/18 22:57, 10F
→
08/18 22:59, , 11F
比方說很難成立這個rating是linear,等距的
08/18 22:59, 11F
→
08/18 23:00, , 12F
社會科學研究實務常是在工具限制下盡量做推論
08/18 23:00, 12F
→
08/18 23:03, , 13F
因此reviewer或許覺得硬把一個不完美的工具說成是完美的
08/18 23:03, 13F
→
08/18 23:03, , 14F
是有點不合
08/18 23:03, 14F
→
08/18 23:05, , 15F
另，社會科學的興趣在統計背後的假設，而非統計本身
08/18 23:05, 15F
→
08/18 23:05, , 16F
如果因為工具限制或資料分布不理想就不做了
08/18 23:05, 16F
→
08/18 23:06, , 17F
似乎是因噎廢食。質性方法、量化方法並存的意義我想在這
08/18 23:06, 17F
→
08/19 00:38, , 18F
我接受bmka的說法,但是實際上likert scale在社科研究裡常用
08/19 00:38, 18F
→
08/19 00:39, , 19F
當然likert scale有其限制
08/19 00:39, 19F
→
08/19 00:41, , 20F
There is nothing wrong with fitting a regression model
08/19 00:41, 20F
→
08/19 00:42, , 21F
to Likert scale data, if the sample size is large enough.
08/19 00:42, 21F
→
08/19 00:43, , 22F
Review算是有點挑語病, 但是作者也要對自己寫的東西負責
08/19 00:43, 22F
→
08/19 00:44, , 23F
Reviewer
08/19 00:44, 23F
→
08/19 00:45, , 24F
我知道要負責&sample size夠大
08/19 00:45, 24F
→
08/19 00:46, , 25F
因為那段文字讀起來的感覺是作者只是follow其他人做的分析
08/19 00:46, 25F
→
08/19 00:46, , 26F
卻不是真的懂自己用了什麼統計方法
08/19 00:46, 26F
→
08/19 00:47, , 27F
受教了,謝謝
08/19 00:47, 27F
→
08/19 00:49, , 28F
寫paper有時這些小細節要小心 :)
08/19 00:49, 28F
→
01/02 14:57, 7年前 , 29F
因此reviewer或 http://yofuk.com
01/02 14:57, 29F
‣ 更多 apson 的文章
文章代碼(AID): #1AYc6F_n (Statistics)
‣ 更多 apson 的文章
文章代碼(AID): #1AYc6F_n (Statistics)