SPSS Statistics

 View Only
  • 1.  Matching subjects into cases and controls

    Posted Thu February 22, 2024 03:27 PM

    Hey, 

    I need to match subjects into cases and controls based on age and gender but also diameter. Cases should have a diameter above a specific value and controls below that value. How do I manage to match cases and controls based on diameter as well as age and gender? I am grateful for detailed answers and step by step explanation as my knowledge in spss is limited. 


    Thanks in advance! 



    ------------------------------
    Adam Ad
    ------------------------------



  • 2.  RE: Matching subjects into cases and controls

    IBM Champion
    Posted Thu February 22, 2024 04:27 PM
    You can construct a matched dataset with specified tolerance (fuzz) values using FUZZY (Data > Case Control Matching).  You would specify the tolerances for age and gender - perhaps zero for both.  Alternatively,  FUZZY allows for a matching function to be supplied that can implement a more complex matching rule.  However, the simplest way to do this, if I understand your requirement, would be to construct two datssets - one of cases and one of controls.  Then in each dataset delete (select out the cases and controls that satisfy the diameter constraints.  Then do the match on age and gender.

    The dialog box help explains the matching part, but the syntax help (enter FUZZY, put the cursor on the command press F1) has details on custom functions.



    --





  • 3.  RE: Matching subjects into cases and controls

    Posted Thu February 22, 2024 05:59 PM

    Thanks alot for the answer! 

    I tried to follow your steps and perform case-control matching by choosing Data>Case-control matching. Then I chose age and gender as "Variables to Match on". I chose Subject ID as Match ID.

    I chose Case-control as "Group indicator" (I sorted cases and controls as 1 for cases and 0 for controls) then I entered 1 0 in "Tolerance factor" square. I entered random names in Match ID Variable and Name for matchgroup. When I ran the command I got error message stating "[errLevel 25] Duplicate variable name." 

    I tried to run the command again without entering a tolerande factor but then I got very weird matching in the new variable column with matches like this one "3846568066704177700.00" which I don't understand what It refers to. In the output viewer it says I got 0 matches. 

    I have spent hours trying to match I hope you can help me figure out how to finally match the cases and controls to be able to perform the analysis. 


    Thanks in advance.



    ------------------------------
    Adam Ad
    ------------------------------



  • 4.  RE: Matching subjects into cases and controls

    IBM Champion
    Posted Fri February 23, 2024 09:46 AM
    I need to see your syntax (via Paste) to understand what you have done.  Note that the output variables cannot already exist in the dataset.  (Send to jkpeck@gmail.com).

    Also, your match variables are age and gender, but  you listed the match values as 1 0.  That seems backwards.  Shouldn't it be 1 0?

    --





  • 5.  RE: Matching subjects into cases and controls

    Posted Fri February 23, 2024 12:51 PM

    Thanks alot for the answer! I was finally able to match the cases and controls. Yes you right! I listed it correctly when I ran the matching. 

    I have matched 1 case with 1 control and I had 5 unmatched cases with the 1 year tolerance factor for age. Now I want to calculate the odds ratio for cases and controls for different risk factors and I wonder how to do it for matched cases and controls for all the 100 subjects (50 cases and 50 controls)? Should I calculate the odds ration for each matched pair or should I do it on group level by calculating the odds ration for all the cases and controls? Now that I have matched pairs It seems as I should count the odds ration for each matched pair but I have no idea if it's correct to do so and how to calculate it? 

    Thanks alot in advance for answer! 



    ------------------------------
    Adam Ad
    ------------------------------



  • 6.  RE: Matching subjects into cases and controls

    IBM Champion
    Posted Sat February 24, 2024 09:39 AM
    You could just run a logistic regression on the selected cases and controls.  The exponentiated coefficients would be the overall odds ratios.

    --