Creating custom attributes for "filter sets" Newbie type question
Locked
Creating custom attributes for "filter sets" Newbie type question
 
|
Although I have been using SPSS since 1972, I am a newbie wrt custom
attributes I am wondering what the expereince of list memers has been about custom attributes. I have a custom attribute MR.Set for which multtiple response set a variable belongs to. It looks ok so far. I am considering creating (a) custom attribute(s) to define * filter sets.* The idea behind this is that there is a known reason for the missing value in other variable when a previous variable has a known answer. For example, OwnCar "Do your own a car?" could be a* filter* variable *Fltered* questions might be Make, Model, Year, etc. when there is a blank answer for Make, Model, Year, etc. a missing value with label of "legitimate skip" could be assigned if the answer to the filter question is "no". Other blanks would be assigned a missing value with a label of "missing for unknown reason". One approach is general FilterING variable. Explains missing values in other variables FilterED variable. Missing values explained by other variable? A second approach gives more detail about each set 2 custom attributes "filter for" with entries like "Make, Model, Year, etc." and "filterED by" with entries like "OwnCar" A third approach would have 2 custom attributes. Fliter.Set with a name for the set like "Car Ownership" Filter.Role with entries like "filter for Make, Model, Year, etc." "FiltererED by OwnCar" ----- 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 |
|
I am not clear on what problem you are trying to solve. In your car example, why would you need more than an NA missing value for the Make, Model, Year questions? You could certainly create a custom attribute that identifies the gating question or, if there are multiple gating questions it could be an array attribute, but what would you want to do with this extra information? On Sat, Oct 17, 2020 at 9:42 AM Art Kendall <[hidden email]> wrote: Although I have been using SPSS since 1972, I am a newbie wrt custom |
Re: Creating custom attributes for "filter sets" Newbie type question
|
|
In reply to this post by Art Kendall
Another idea: You could use an attribute "tabletype" to indicate a strategy for how to report the results of a variable. Something like a) T01 = Frequencies only b) T02 = Frequencies + Top 2 Box c) T03 = Frequencies + Low 2 Box d) T04 = Frequencies + Mean e) T05 = None Mario Giesel Munich, Germany ---
Am Samstag, 17. Oktober 2020, 17:42:26 MESZ hat Art Kendall <[hidden email]> Folgendes geschrieben:
Although I have been using SPSS since 1972, I am a newbie wrt custom attributes I am wondering what the expereince of list memers has been about custom attributes. I have a custom attribute MR.Set for which multtiple response set a variable belongs to. It looks ok so far. I am considering creating (a) custom attribute(s) to define * filter sets.* The idea behind this is that there is a known reason for the missing value in other variable when a previous variable has a known answer. For example, OwnCar "Do your own a car?" could be a* filter* variable *Fltered* questions might be Make, Model, Year, etc. when there is a blank answer for Make, Model, Year, etc. a missing value with label of "legitimate skip" could be assigned if the answer to the filter question is "no". Other blanks would be assigned a missing value with a label of "missing for unknown reason". One approach is general FilterING variable. Explains missing values in other variables FilterED variable. Missing values explained by other variable? A second approach gives more detail about each set 2 custom attributes "filter for" with entries like "Make, Model, Year, etc." and "filterED by" with entries like "OwnCar" A third approach would have 2 custom attributes. Fliter.Set with a name for the set like "Car Ownership" Filter.Role with entries like "filter for Make, Model, Year, etc." "FiltererED by OwnCar" ----- Art Kendall Social Research Consultants -- ===================== 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 |
|
|
Along these lines, for repeating or similar surveys or other data sources, I have advocated that in the data collection or extraction step, that variables be classified by area - demographic, economic, etc and type of analysis, and then the analysis/reporting code could use those attributes to create the report(s). That way, the code could accommodate changes in the survey or database without modifications or at least with fewer changes. Of course, one would need to write the syntax-generating rules in Python code taking into account both the built-in metadata and the custom attributes.
|
Re: Creating custom attributes for "filter sets" Newbie type question
|
|
I did create a custom attribute called Var.Grouping for what you called area
with entries like "case tracing for followup" "Demographic", "sanitation", "Nutrition", "Hygiene" and I'll look into specifying how to report. It might work in this situation. In clinical trials and other experimental and quasi-experimental research, ther is a growing pressure to specify analysis methods before proposal are approved. ----- 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 |
Re: Creating custom attributes for "filter sets" Newbie type question
|
|
In reply to this post by Jon Peck
SOAPBOX on.
The tangled messy data set I received shows why people wanting to do research in social and behavioral science areas should seek guidance from people with backgrounds and experience in the data science end of social and behavioral science. Humanitarian NGOs can seek volunteers from Universities, Statistics Without Borders, or others. Human Rights NGOs can seek volunteers from the AAAS On-Call Scientists, Universities, Statistics Without Borders, or others. SOAPBOX off. The reasons I want to document filter/gating variables include 1) the NGO people who work with the data in the future may not have SPSS available. They would need information that documents what is in the .SAV and XLSX versions. 2) if the NGO gathers more data in future the more complete data definition would help them in the design of data-gathering instuments. "NA" or "Not applicable" would be more explicit "legitimate skips". 3) clarifying reasons for missing data to help substantive interpretation. 4) to aid me in figuring out whether I can create internal SPSS data validation rules. 5) If I can minimize unexlained missing data, there is a possibility that something meaningful can be done with the data. The received data has vast amounts of missing data. Zero's for numeric and blank for strings. Zeros start off labelled as "possibly ambiguous zero". Why is a person' age coded zero? After autorecode, the string variables with blank are labelled -1 "missing for unknown reason". The goal of figuring out which variables are filter variables is to aid in changing some zeros or -1's to to things like "NA legitimate skip" "all HH this city are missing" "data entry or field error". Some -1 missing values in Multiple Response can be "cured" by something like"no mentions in MR set" or "ran out of mentions" Cases with anomolus values or "data entry or field error" for example may be reasons to see which interviewer is having difficulties with the task or the language of teh NGO's clients. ----- 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 |
|
|
If you are not already using it, you might want to take advantage of the ADD DOCUMENT (Utilities > Data File Comments) or DOCUMENT command and, for programmatic use, the DATAFILE ATTRIBUTE command. These properties are included in SAV files. On Sat, Oct 17, 2020 at 2:05 PM Art Kendall <[hidden email]> wrote: SOAPBOX on. |
Re: Creating custom attributes for "filter sets" Newbie type question
|
|
After discussions and consideration, I decided to go with 2 attributes.
The first attribute I'll call Filter.Set The entry for a variable would have contents that reflect meaning like "OwnCar" "OwnHouse" The second attribute would have values like '1' for the filtering variable and 2, 3 etc for order in the set The Filter.Sets contain small numbers of variables so the name of the set should carry enough information. Then I would rely on the SORT in SPSS being stable. Some syntax might look like this SORT VARIABLES BY ATTRIBUTE Var.Grouping. SORT VARIABLES BY ATTRIBUTE MR.Set. SORT VARIABLES BY ATTRIBUTE Filter.Set. SORT VARIABLES BY ATTRIBUTE Filter.Order. ----- 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 |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |