Re: CATPCA "Impute Extra" category
Locked
Re: CATPCA "Impute Extra" category
 
|
At 12:56 AM 1/28/2009, Kooij, A.J. van der wrote:
> >It occurred to me that these DK/NA responses did not necessarily fit on > the otherwise ordinal scale ... > >CATPCA has the missing option Impute Extra category. The quantification >for the extra category is free, i.e, not restricted according to the >optimal scaling level for the variable. So, if you specify the dk/na >categories as missing and choose Impute Extra category, the quantified >dk/na categories will not be ordinally restricted (if you have categories >1 to 5 and specify the middle category as missing, CATPCA will impute 6 >for the middle category. The quantified values of categories 1, 2, 4 and 5 >will be ordinally related to the original category values, but the >quantified value of category 6 does not need to be higher than the >quantified value of category 5). Thank you for this suggestion, which I now want to attempt to implement. You wrote, >" if you specify the dk/na categories as missing..." So I do this from the SPSS Variable View? For example, if my dk/na category is coded "3", I specify "3" as one of the "discrete missing values"? I assume there must be a way to do the same thing with a syntax command, as well. Then in my CATPCA command sequence, if I have variables E1 E2 E3 E4 that I want to treat this way, is this the right syntax? /MISSING = E1 E2 E3 E4 (Extracat) (On a related note, I am not clear on the difference between ACTIVE and PASSIVE in these /Missing commands.) And importantly, do I understand correctly that if I do this, CATPCA will treat ALL missing value codes the same way, not just the dk/na category? For example, will it treat a system missing, such as a blank cell, the same as a "0", and those will be treated the same way as my dk/na code "3" that I have declared missing? Thanks, Bob Schacht Robert M. Schacht, Ph.D. <[hidden email]> Pacific Basin Rehabilitation Research & Training Center 1268 Young Street, Suite #204 Research Center, University of Hawaii Honolulu, HI 96814 ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: CATPCA "Impute Extra" category
|
|
>" if you specify the dk/na categories as missing..." Yes. Yes: MISSING VALUES var1 var2 (3). >>Then in my CATPCA command sequence, if I have variables E1 E2 E3 E4 >>that I want to treat this way, is this the right syntax? No, the right syntax is /MISSING E1 E2 E3 E4 (Active extra). With / ... (Extracat), the missing handling is Passive (the default if active or passive is not specified). The imputation of an extra category if the missing option is Passive is only for computing the correlations between the transformed variables if /PRINT ... corr, which is done after the solution is computed; the transformed variable is imputed with the quantified value for an extra category (or the quantified value of the mode category if /MISSING ... (Passive mode)). >>(On a related note, I am not clear on the difference between ACTIVE >>and PASSIVE in these /Missing commands.) ACTIVE: impute missings. PASSIVE: no imputation and no deletion, but exclusion of missing values. See below for more details about PASSIVE.
Unfortunately, yes. So, if a variable has missing values this approach can not be used (unless you would assume that people with dk/na answer have something in common with people for which answer is missing). In that case, to see if the value of 3 for the dk/na category fits in an ordinal scale, choose nominal scaling level in stead of ordinal and look at the transformation plots. Regards, Anita Passive handling of missings: With the missing option Passive, the missing entries on a variable are not imputed but excluded in the analysis. This option is something different than Listwise or Pairwise exclusion. With Passive missing treatment nothing is excluded; all valid values are in the analysis. This strategy is possible in CATPCA because the solution is not found from the correlation matrix (as in standard PCA), but from the data itself. (In the CATPCA algorithm passive missings are handled by applying case weights in the computation of the component scores; the weight for a case is the number of variables with a valid value for that case.) If Passive missing is applied, the loadings are not the correlations between the variables and the component scores (so you could obtain loadings with absolute value greater than 1) and % of VAF is undefined (that is why Model Summary table displays VAF, not %VAF if Passive missing for one or more variables is used). Also, the transformed data will have missings where the original data have missing, which is a disadvantage if you want to use the transformed data for further analysis with different analysis methods, because in other methods the Passive missing option is not available. So, for interpretation and further analysis it is more convenient to impute missing values or to choose listwise deletion. To see the influence of deleting cases or imputation of missings, you can compare results to the results when applying Passive option; if the results do not differ much, deletion or imputation does not have much effect and thus is a valid strategy to handle missing values. From: Bob Schacht [mailto:[hidden email]] Sent: Fri 17-Apr-09 22:19 To: Kooij, A.J. van der; [hidden email] Subject: Re: CATPCA "Impute Extra" category At 12:56 AM 1/28/2009, Kooij, A.J. van der wrote: ********************************************************************** This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. **********************************************************************
|
| Free forum by Nabble | Edit this page |