Dear all,
I cannot find the scoring equations for the CATPCA algorithm... I ran CATPCA on a set of ordinal variables and need to calculate the scores of future cases by implementing the CATPCA scoring equations in "Transform", "Compute variable" on the basis of calculated loadings. Thank you very much in advance for your information! Paolo ===================== 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 |
Have you looked up CATPCA in the Algorithms documentation? On Fri, Dec 27, 2019 at 1:08 PM Paolo <[hidden email]> wrote: Dear all, |
In reply to this post by Paolo
What kinds of constructs are you exploring? Attitudes, values, personality,
moods, abilities, knowledge, etc.? Please clarify what the response scale is on your items. I.e., what are legitimate values and what are the labels for those values. Why do you believe that the response scale is severely discrepant for interval level? It is over 50 years since it was well established the unit weights are preferable in summative scales. So after being sure that (1) you are not retaining too many factors (2) items load cleanly on only one factor (3)items are recoded in a consistent direction simply sum or average those subsets of items. Did you do a principal factors analysis and parallel analysis to assure that you did not retain to may factors? In that did you use varimax rotation to assure there is divergent validity among the constructs underlying the factors? If this is a pre-existing set of scales, what scoring keys were used previously? ----- Art Kendall Social Research Consultants -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== 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
Art Kendall
Social Research Consultants |
Hello,
On 02/01/2020 20:33, Art Kendall wrote: > It is over 50 years since it was well established the unit weights are > preferable in summative scales. Can you give some info where to find authors who establish that finding ? Thanks in advance - ftr ===================== 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 |
Sorry, I donated all my files and books when I retired in 2001. These days
library systems have full test search systems for many journals. Look for articles from the journal /*Psychometrika*/* in the 50s and 60s with the terms "factor analysis" and "unit weights". How similar/ different are the scoring keys when you use CATPCA and when you use the conventional FACTOR with Principal Axes and Varimax rotation? It helps list members give advice we can stand by when we have a fuller grasp of the circumstances in which you are using factor analysis. What kind of constructs are you looking to create summative scales for? What is the response scale for items you think might go together? How large is the pool of candidate items? How many cases are in your data set? Why do you think the item response scale is severely discrepant from interval level? ----- Art Kendall Social Research Consultants -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== 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
Art Kendall
Social Research Consultants |
The references listed in the algorithms manual for CATPCA may be of help. On Sun, Jan 5, 2020 at 4:26 PM Art Kendall <[hidden email]> wrote: Sorry, I donated all my files and books when I retired in 2001. These days |
In reply to this post by Paolo
Consider SUPPLEMENTARY OBJECTS subcommand
27.12.2019 11:40, Paolo пишет: > Dear all, > I cannot find the scoring equations for the CATPCA algorithm... > I ran CATPCA on a set of ordinal variables and need to calculate the scores of future cases by implementing the CATPCA scoring equations in "Transform", "Compute variable" on the basis of calculated loadings. > Thank you very much in advance for your information! > > Paolo > > ===================== > 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 > ===================== 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 |
Dear Kirill,
Thank you for your information.
I checked the SUPPLEMENTARY OBJECTS command but this command does not output the scoring formula used to calculated the scores of extra cases (it only outputs the score values of supplementary objects in the SPSS file).
Given that these extra cases will be available at a later date and in a separate data set, it would be best to know the formula for scoring them rather than merging/adding these cases to the original data set and rerun the CATPCA specifying these additional
cases as supplementary objects to score.
Thank you very much for any additional advice.
Kind regards,
Paolo
Paolo Piaggi, PhD
Obesity and Diabetes Clinical Research Section NIH/NIDDK 4212 N 16th St. Phoenix, AZ 85016 From: Kirill Orlov <[hidden email]>
Sent: Monday, January 6, 2020 04:09 AM To: Piaggi, Paolo (NIH/NIDDK) [C] <[hidden email]>; [hidden email] <[hidden email]> Subject: Re: CATPCA equations for scoring Consider SUPPLEMENTARY OBJECTS subcommand
27.12.2019 11:40, Paolo пишет: > Dear all, > I cannot find the scoring equations for the CATPCA algorithm... > I ran CATPCA on a set of ordinal variables and need to calculate the scores of future cases by implementing the CATPCA scoring equations in "Transform", "Compute variable" on the basis of calculated loadings. > Thank you very much in advance for your information! > > Paolo > > ===================== > 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 > |
for the sake of presentation, let us assume that you have actual Likert
items. which are rarely severely discrepant from the interval level. Let us also assume that -- you have at least 10 cases per potential scale item; -- you found 2 factors with items that loaded cleanly and were reasonably coherent in meaning. value labels MyItem01 to MyItem50 1 'Strongly Disagree' 2 'Disagree" 3 'neither agree nor disagree' 4 'Agree 5 'Strongly agree'. Also some items needed to be reflected because they loaded negatively. Scoring would be something like this **** Scoring new cases ****. Recode MyItem03, MyItem07, MyItem13, MyItem42 into , MyItemR13, MyItemR42 (5=1)(4=2)(3=3)(2=4)(1=5). Compute MyConstruct1 = Mean(MyItemR03, MyItemR07, MyItem01, MyItem33, MyItem41, MyItem50). Compute MyConstruct2 = Mean(MyItemR13, MyItemR42, MyItem11, MyItem22, MyItem28, MyItem49). **** End Scoring new cases ****. This approach puts the construct scores on the same response scale as that underlying the item response scales. Some disciplines use total scores. If so use "Sum" instead of "Mean" above. Also unless you have millions of cases, or it is a long time until there are new cases, it is often fairly easy to redo the whole analysis. Of course, the context in which the questions come up, it critical to suggesting a more detailed approach. ----- Art Kendall Social Research Consultants -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== 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
Art Kendall
Social Research Consultants |
The second line of the Recode command was incomplete.
Recode MyItem03, MyItem07, MyItem13, MyItem42 into MyItemR03,MyItemR07,MyItemR13,MyItemR42 (5=1)(4=2)(3=3)(2=4)(1=5). ----- Art Kendall Social Research Consultants -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== 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
Art Kendall
Social Research Consultants |
In reply to this post by Paolo
Just append the new gathered cases to the old dataset (which you
must keep) and specify those cases as supplementary objects.
06.01.2020 16:34, Piaggi, Paolo
(NIH/NIDDK) [C] пишет:
===================== 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 |
Free forum by Nabble | Edit this page |