I have completed the 1:1 case control matching process in SPSS and have successfully got the file with just the cases and controls in it (each participant on a separate row).
Theoretically I could run an independent t-test but is this incorrect as the data is matched?
If an independent t-test is wrong, does anyone know how to restructure the data further so that each control is on the same row as the case, in order to allowed a paired t-test?
Many thanks
Dear Andrea,
It is not defficult. Only put the control cases and treatments cases in column speartaly. Seee https://youtu.be/MJGk2sg4EZU
You need to use CASESTOVARS to restructure the data from LONG to WIDE. Here is an example using some made up data. You should be able to tweak it to fit your situation. HTH.
* Generate some fake data to illustrate how to restructure the file.
NEW FILE.
DATASET CLOSE ALL.
DATA LIST List / MatchPair (F2.0) Case (F1).
BEGIN DATA
1 1
2 1
3 1
4 1
5 1
1 0
2 0
3 0
4 0
5 0
END DATA.
COMPUTE Y = RV.NORMAL(50,10).
LIST.
SORT CASES BY MatchPair Case.
CASESTOVARS
/ID=MatchPair
/INDEX=Case
/GROUPBY=VARIABLE.
LIST.
Reply to Ahmed and Bruce: Thanks so much for your help (and syntax!) but the issue is not in how to run the paired t-test itself, the issue is how to get the current dataset in the correct form for it. Unfortunately as we have thousands of rows we can't just manually paste the results from the controls into the same row as the cases. Sorry if I should have made that clearer.
As suggested by Soheil I have included below an example of the way the data is structured (I can't include the original data):
ID Age Gender BMI MatchID
101 31 0 25 104
102 25 1 27 105
103 27 0 22 106
104 54 0 24 101
105 21 1 21 102
106 21 0 45 103
...and so on for hundreds/thousands of rows
How can I move the controls on the same row as it's case? So for example I can do a paired t-test for BMI? For example how can I get subject 104 on the same row as subject 101 without cutting and pasting it? (I have too many rows for that to be feasible).
Dear Andrea L. Darling,
it is very easy go to Data and then split file then select compare group , put the third variable that includes codes e.g.0 1 0 1 1 1 1 2 2 2 2 2 0 0 00 0 0 01 1 1 1 11 then you copy each group speratly and put in a new column.
My regardsm
Hello Andrea L. Darling. Thank you for providing an example of how the data is structured. It strikes me as an odd way to have coded the data. It is more common for a variable called MatchID to have the same value for both members of a matched pair. Using your example of 101 and 104 being paired, I would have coded it like this:
ID MatchPair
101 1
102 2
103 3
104 1
105 2
106 3
But presumably, that ship has sailed, and you have to figure out a way to do the data management with the data the way you have it!
Q. Taking your example of IDs 101 and 104 being paired, how can you tell which one is the case and which is the control? Do you have some rule such when ID < MatchID, ID is the case and MatchID is the control?
Thank you for clarifying.
Meanwhile, here is a way that will sort the records you have shown so that the two rows for a match pair are listed consecutively. Perhaps that will get you part way to where you need to be.
*** Start of syntax ***.
NEW FILE.
DATASET CLOSE ALL.
DATA LIST List / ID Age (2F5.0) Gender (F1) BMI MatchID (2F5.0).
BEGIN DATA
101 31 0 25 104
102 25 1 27 105
103 27 0 22 106
104 54 0 24 101
105 21 1 21 102
106 21 0 45 103
END DATA.
COMPUTE LowHigh = MIN(ID,MatchID)*1000 + MIN(ID,MatchID).
FORMATS LowHigh (F8.0).
SORT CASES BY LowHigh.
LIST.
*** End of syntax ***.
Result:
ID Age Gender BMI MatchID LowHigh
101 31 0 25 104 101101
104 54 0 24 101 101101
102 25 1 27 105 102102
105 21 1 21 102 102102
103 27 0 22 106 103103
106 21 0 45 103 103103
Here's some more syntax that may help. As noted in the comments, I have made an assumption about which rows represent cases and which rows represent controls. That part will need to be modified if my guess was wrong.
*** Start of syntax ***.
NEW FILE.
DATASET CLOSE ALL.
DATA LIST List / ID Age (2F5.0) Gender (F1) BMI MatchID (2F5.0).
BEGIN DATA
101 31 0 25 104
102 25 1 27 105
103 27 0 22 106
104 54 0 24 101
105 21 1 21 102
106 21 0 45 103
END DATA.
***************************************************.
* NOTE: I am assuming Cases appear on rows where
* ID < MatchID. Andrea will have to clarify how
* Cases & Controls are actually identified.
COMPUTE Case = ID < MatchID.
FORMATS Case (F1).
***************************************************.
COMPUTE LowHigh = MIN(ID,MatchID)*1000 + MIN(ID,MatchID).
FORMATS LowHigh (F8.0).
SORT CASES BY LowHigh.
LIST.
* Generate a fake dependent variable, as none was given.
COMPUTE Y = RV.NORMAL(50,10).
LIST.
SORT CASES BY LowHigh Case.
CASESTOVARS
/ID=LowHigh
/INDEX=Case
/SEPARATOR=""
/GROUPBY=VARIABLE.
LIST.
*** End of syntax ***.
Bruce Weaver Ahmad Alshallawi Thanks for your replies!
Bruce's first syntax has worked- it has got the data to the point (cases and controls in rows, one below each other) where we can move forward to get a successful result. Thanks for the second syntax but unfortunately it wont work for us as in the real example ID is not always lower than MatchID.
Thanks to everyone who has replied, as I say we can now move forward with the analysis which is great.
p.s. yes it is odd- SPSS labels each case with the ID of the control (and each control with the ID of the case) rather than labelling each member of the case and control with same number.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS