[求助]关于因子分析的原始数据

S

stream

普通会员
大家好！我想对65个变量的频率进行因子分析，比如what, how, which在locness 中出现的频率分别是51，42，14.............,共有65个类似的变量和各自的频率， 要对这些数据进行因子分析，在spss中要输入的原始数据是什么呢？我试过把这65个变量的频率输入，然后analyze--data reduction---factor analysis但是结果输出中的相关系数矩阵提示错误！请各位高手指点一下！谢谢！
 
这样是不能进行因子分析的。

因子分析的目的是从大量的显性变量中简化数据，分析出其中的隐性变量。

你说的"类似的变量"值得是什么？可以从这些"类似的变量"所可能隐含的共性来着手。
 
Biber used Factor Analysis to bring down 67 linguistic features into 6 dimensions or factors. It appears that you have accepted Biber's dimensions and what you really want to do is compute the dimension scores (factor scores) of the genres in your study. Unit C5 "Conversation and Speech in American English" in the book Corpus-Based Lanuage Studies" (Routledge London/New York, 2006) shows you step by step how to start from here. The following paper shows the result of this analysis:
http://eprints.lancs.ac.uk/46/
But you will need to consult the book to see how this kind of analysis is done.
 
but how can i find this book you mentioned,Dr.Xiao? As to the paper you showed to me, i cannot see the body.
 
Here is the book: http://www.routledge.com/textbooks/0415286239/default.asp

Click on the PDF sign at the Eprint site to read the full paper.
 
谢谢您！但是我还是有个问题。因子分析的原理好像是把多个变量减少为少数几个变量，它要求的原始数据应该不仅仅是每个语言特征在语料库中出现的总频率， 而应该是在很多个文章中出现的不同频数，对不对？ 但是，事实上如果用locness， 它是一个整体，检索出的结果只有一个总的频数，这样的话， 原始数据应该是什么呢？谢谢！[emq1][emq1]
 
Also find the frequency of each linguistic feature in each corpus file in a corpus/register (use the dispersion function in WordSmith), then compute the minimum, maximum, mean, range and std deviation for each feature in that corpus/genre (copy the left panel in the Wordsmith dispersion window into SPSS) before you can compute dimension/factor scores.
 
回复：[求助]关于因子分析的原始数据

以下是引用 stream 在 2006-4-5 15:26:02 的发言：
many thanks,Dr.Xiao. But i didn't find the Wordsmith dispersion function yet.[emq10]
Click to expand...


you may search "dispersion" in Wordsmith Tools Help.
 
There is a button for (dispersion) Plot on the toolbar (with the letter p on the icon) when you have made a concordance.
 
在对频数因子分析时，我在进行描述性统计时已经把绝对频数一一输入了spss进行了计算，能不能把这些已经输入的绝对频数直接进行标准化处理作为原始数据呢？如果不行，是不是就应该是把WSM中的dispersion plot显示的normalized frequency一一输入spss作为原始变量，这就是说我要把绝对频数和相对频数都要分别输入，才能进行因子分析吗？
 
You must log in or register to reply here.
Back
顶部