Date: Mon, 15 Apr 1996 09:45:22 -0400
Reply-To: Claire Durand <durandc@ERE.UMONTREAL.CA>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: Claire Durand <durandc@ERE.UMONTREAL.CA>
Subject: Re: scale development
In-Reply-To: <960413230301.20622564@MARY.FORDHAM.EDU>
On Sat, 13 Apr 1996, Shawn Utsey wrote:
>
> in completing my scale participants are directed to indicate their
> reaction to a given stressful event, if the event has never happened they
> should also indicate this. the scale is as follows: 0=never happened,
> 1=happened, but not upsetting, 2=happened and slightly upsetting, 3=
> happened and upset, and 4 = happened and extremely upset.
>
> after collecting data based on the above response options i entered the
> data into spss for windows and conducted a principle components factor
> analsis. i entered the data as is. the feedback that i have received is
> that i should have gotten rid of the "0" because they indicate that the
> event did not happen. according to the reveiwer by including the "0" i
> create error variance and create a non-linear distribution.
>
> can someone tell me how you can get rid of the "0" and not create a
> missing varible, which will cause cases to be excluded. also because
> it is not a missing variable, but part of the scoring scheme.
>
>
If you don't want to lose people, the only solution that could be
accepted would be to declare the "0"" as missing and then use the option
"missing=mean substitute" . The rational for this is that you consider
that the "0" are equivalent to "not applicable" i.e. they cannot have a
reaction to something that never occured to them. In substituting the
mean, you adopt a "heuristic solution" because those cases who are
attrributed the mean (I hope they are not too many) will not influence
the factor solution.
Hope this helps...
Claire Durand e-mail: durandc@ere.umontreal.ca
Universite de Montreal
| 
