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 (August 2004)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Tue, 17 Aug 2004 08:22:27 -0300
Reply-To:     Hector Maletta <hmaletta@fibertel.com.ar>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Hector Maletta <hmaletta@fibertel.com.ar>
Subject:      Re: Data
Comments: To: Claudia Beinert <claudia.beinert@confirm-md.de>
In-Reply-To:  <003301c4842e$72fef4d0$2101a8c0@ConFiRM3>
Content-Type: text/plain; charset="iso-8859-1"

Claudia: Thanks for the clarification as to the structure of your data. To test for a normal distribution of frequencies you must have an interval-level variable, whereas your variable (rating grade) is a categorical ordinal variable. The categories have an intrinsic order, but no quantitative value. You may, of course, assign the categories some set of arbitrary values, such as making aaa=1, aa=2, a=3, bbb=4 and so on, but with your data you have no basis for such transformation. Any other set of numeric values would be equally valid (aaa=1, aa=8, a=9, bbb=19 and so on). Therefore, what you have is the distribution of issuers over an ordinal variable with a number of ordered qualitative categories without numeric value. This cannot have a normal (or other) mathematical distribution because these distributions are defined for a quantitative or interval-level variable (such as income or age) and not for categorical ordinal variables such as rating grade. For a more impresionistic judgment SPSS may produce an histogram graph and superimpose a normal distribution. This procedure assumes the categories are equally spaced, which is not necessarily so, but may give you an idea. To produce such a graph you may use the CHART menu, or the graph options in FREQUENCIES, and later edit the chart to add the normal curve.

Hector

> -----Original Message----- > From: Claudia Beinert [mailto:claudia.beinert@confirm-md.de] > Sent: Tuesday, August 17, 2004 4:47 AM > To: Hector Maletta > Subject: Re: Data > > > Hi Hector, > > (a) the Structure of my data is exactly as I wrote: All I > have, are the values of the frequency distribution. Therefore > I know, that 60 Issuers have the rating grade AAA, 200 > Issuers have AA...and so on. All the 2.000 Issuers with the > Rating grades from AAA til C are one Column. Every new > Column is a new year in my data file. My problem is a data > problem: I actually need the data in that form (table 1), > that SPss can test for a certain form of distribution: > > No.Issuer Rating grade > 1 aaa > 2 bb > 3 c > 4 aaa > 5 a > 6 aaa > 8 aaa > 9 bb > 10 aaa > > > But all I have is (table 2): > > Rating Grade Frequency > aaa 5 Issuers > a 1 Issuer > bb 2 Issuer > c 1 Issuer > > > My question: is there a possibility, that spss can take the > data as in table 2 to test for a distribution..or do i have > to convert them at first, that they look like in table 1. > > (b) I chose Kol-Smirnow, because I want to test, if the > freqency -distribution for all Issuers over the rating grades > are normally distributed. So the fixed value I am testing ist > the normal distribution. If I already know that the > distribution is normal, i can than take the one-sample-t-test > to test for a fixed value, but that's not what I want. > > Thanks for your help > > Kind regards, > Mit freundlichen Grüßen > > Claudia Beinert > _______________________ > Leiterin Produktentwicklung > ConFiRM GmbH > Hegelstr. 18 > 39104 Magdeburg > > Tel. ++49(0) 391/ 254 9997 > Fax. ++49(0) 391/254 9999 > ----- Original Message ----- > From: "Hector Maletta" <hmaletta@fibertel.com.ar> > To: "'Claudia Beinert'" <claudia.beinert@confirm-md.de> > Sent: Monday, August 16, 2004 5:47 PM > Subject: RE: Data > > > > Claudia, > > Your message does not make clear (a) the structure of your > data file > > and > (b) > > what exactly you want to test. > > > > Regarding 1: You may have a data file with one variable for each of > > the categories (i.e. one column for AAA, one column for AA, one for > > BBB, > etc.), > > or you may have one single column for the category (with > values AAA, > > AA, > A, > > BBB, BB etc.) or you may have one column for the category cited as > > "first" (AAA, AA, A), one column for the second (BBB, BB, B) and so > > on. > > > > Regarding 2: what do you mean by "testing it with one-sample > > Kolmogorov-Smirnov"? Against which you'd test your data? > You may want > > to know whether the frequency for AAA is significantly > different from > > the frequency for BBB, and so on for every pair of categories (too > > many > pairs), > > or you may want to test whether the observed frequencies as a whole > > are different from the expected random frequencies (which should be > > uniform > for > > all categories, I suppose), or what else? > > > > By the way, one-sample tests indicate whether an observed > measure (for > > instance, an average for an interval variable) is significantly > > different from a value arbitrarily fixed by the researcher (for > > instance, whether > the > > average income in your sample is significantly different from > > $20,000). > This > > arbitrary external value may come, for instance, from some other > > research > or > > from national accounts or census data, or whatever you want > to compare > your > > data with. From your question it is no0t clear why you should be > > applying this kind of test. > > > > So you better clarify, to see whether somebody can help. > > > > Hector > > > > > -----Original Message----- > > > From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On > > > Behalf Of Claudia Beinert > > > Sent: Monday, August 16, 2004 12:31 PM > > > To: SPSSX-L@LISTSERV.UGA.EDU > > > Subject: Data > > > > > > > > > Hi, > > > > > > I would like to test a Frequency from the rating industry like > > > > > > AAA : 60 Issuers > > > AA : 200 Issuer > > > A : 800 Issuer > > > BBB : 500 Issuer > > > BB : 500 Issuer > > > ... : an so on > > > > > > > > > with One-Sample-Kolmog.-Smirnov-Test for Normal distribution. > > > > > > Is there in way in SPss (I am running version 11.0) to > test the data > > > given above without typing in each single issuer? > > > > > > Kind regards an Thanks for your help. > > > > > > Claudia Beinert > > > _______________________ > > > Leiterin Produktentwicklung > > > ConFiRM GmbH > > > Hegelstr. 18 > > > 39104 Magdeburg > > > Germany > > > > > > >


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