LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (September 2005)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Sun, 4 Sep 2005 14:16:30 -0400
Reply-To:     "Frank J. Gallo" <fjgallo@verizon.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Frank J. Gallo" <fjgallo@verizon.net>
Subject:      Re: Combining Variable Scores
Comments: To: Hector Maletta <hmaletta@fibertel.com.ar>
In-Reply-To:  <S475185AbVIDQ2J/20050904162810Z+79527@avas-mr01.fibertel.com.ar>
Content-type: text/plain; charset=US-ASCII

Hi Hector,

Thank you very much for your input. I apologize for not giving a better explanation of my situation.

Respondents rated (1 to 6) the violence severity of 37 different behaviors. I collected 5 samples from the target population. Samples were collected at different time points, but within one year. I do not expect the distributions of the samples to be significantly different -- within the limits of sampling errors. However, this is an empirical question. I would like to combine the several data sets if (a) they do not show large statistical differences between associated distributions and (b) I can document other similarities. The population parameters are unknown. So, my first step is to perform an exploratory data analysis: calculate the mean, median stdev, min, max, Q1 and Q3 statistics for each sample. Then calculate the same statistics after combining the samples. Then do some other procedures such as a graphical analysis, analysis of variance, etc. Any further thoughts are appreciated.

Thanks, Frank

-----Original Message----- From: Hector Maletta [mailto:hmaletta@fibertel.com.ar] Sent: Sunday, September 04, 2005 12:28 PM To: 'Frank J. Gallo'; SPSSX-L@LISTSERV.UGA.EDU Subject: RE: Combining Variable Scores

Frank, What do you mean by "combining" the scores? As you know, there are different ways to do that. One simple way is just obtaining the average or sum of variable scores, but this would give all variables the same weight. A more elaborate way is factor analysis or some variant of it: if all your variables reflect one underlying factor or trait, then the first factor extracted should account for a large portion of total variance in your 37 variables, and the scores for that first factor may be used as a single variable representing the main component of the common variance in your variables.

Once you have your final score for the synthetic variable representing your 37 original variables, obtaining the summary measures you mention is quite easy with the FREQUENCIES or DESCRIPTIVES command. In your case FREQUENCIES is better because you want the quartiles too.

To summarize:

1. Obtain a single variable representing your 37 scores. 1.1. Obtain it as a simple average. COMPUTE MEANSCOR=MEAN(LABEL1 TO LABEL37). If all your 37 variables use the same scale (say, 1 to 5) this may be enough. If they have different ranges and units, you may better standardize them to have zero mean and unit standard deviation. This can be done with the SAVE option in the DESCRIPTIVE command, applied BEFORE the COMPUTE. The SAVE keyword will create 37 new variables named ZLABEL1 to ZLABEL37, which will be the standardized version of your variables. DESCRIPTIVES LABEL1 TO LABEL37/SAVE. COMPUTE MEANSCOR=MEAN(ZLABEL1 TO ZLABEL37). 1.2. Obtain it by means of FACTOR ANALYSIS: FACTOR VARIABLES LABEL1 TO LABEL37/PRINT ALL/SAVE REG FASCOR. This would extract all factors with eigenvalues above 1, and would save the scores to the file under new variables named FASCOR1 to FASCORk (where k is the last factor extracted). In the output look at the VARIANCE EXPLAINED table. Judging from the contribution of the first factor to explaining all variance in the original variables, you may decide whether the contribution of the first factor is much larger than the second and later factors, or perhaps your variables are in fact measuring two or more different underlying factors of similar importance.

2. Once you have a single score, say FASCOR1 or MEANSCOR, you may know the main statistics by using FREQUENCIES:

FREQUENCIES FASCOR1 /format notable/ntiles 25/statistics all.

This would not produce an actual frequency distribution (too many values for that), but will give you the quartiles and all the summary measures you want (and some more).

Hector

> -----Original Message----- > From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] > On Behalf Of Frank J. Gallo > Sent: Sunday, September 04, 2005 11:48 AM > To: SPSSX-L@LISTSERV.UGA.EDU > Subject: Combining Variable Scores > > Hi All, > > > > Still green at writing syntax, and I am hoping that someone > can suggest some syntax for the following run: > > > > -- I have 37 variables (label1 - label37) > > -- sample: n = 50 cases > > -- I would like to combine the variable scores and then > compute the mean, median stdev, min, max, Q1 and Q3 > statistics for the sample (n=50). > > > > Your help is greatly appreciated. > > Frank > > __________ Informacisn de NOD32 1.1208 (20050902) __________ > > Este mensaje ha sido analizado con NOD32 Antivirus System > http://www.nod32.com > >


Back to: Top of message | Previous page | Main SPSSX-L page