Comparing affected sides instead of just left and right.

#1
Hi.
I'm relatively new to statistics and SPSS.

I've been testing the shoulder strength of 200 shoulder patients divided into two groups. I've been testing both of their shoulders, so I could have a reference value.

I would like to compare the affected shoulder of group 1 to the affected shoulder of group 2. The problem is, I entered the data as left and right instead of affected side and unaffected side.

I have entered the affected side for each patients as a categoric value (0=right and 1= left), but I can't figure out how I combine this, so I can compare the affected side of group 1 to the affected side of group 2 - instead of just comparing left and right.
I'm using the independent T-test.

Any help is much appreciated. Thank you.
 
#2
If I've understood correctly, your data looks something like this, right?

Code:
id     grp     left   right  aff_side
1	1	62	23	1
2	2	78	34	0
3	1	35	75	1
4	2	43	54	1
5	1	37	87	0
6	2	53	97	0
If that's correct, then you need to use some if statements to create your DV like so:

Code:
if aff_side = 0 my_dv = right.
if aff_side = 1 my_dv = left.
exe.
Then you can run your comparison with group as the IV and my_dv as the DV.
 
#3
If I've understood correctly, your data looks something like this, right?

Code:
id     grp     left   right  aff_side
1	1	62	23	1
2	2	78	34	0
3	1	35	75	1
4	2	43	54	1
5	1	37	87	0
6	2	53	97	0
If that's correct, then you need to use some if statements to create your DV like so:

Code:
if aff_side = 0 my_dv = right.
if aff_side = 1 my_dv = left.
exe.
Then you can run your comparison with group as the IV and my_dv as the DV.
Thank you. It makes sense. Would you care to elaborate how I do this exactly?
 
#4
Open your dataset then open a syntax window by going to File -> New -> Syntax. Paste the following into the syntax window and change 'aff_side', 'right', and 'left' into whatever you've called those variables in your dataset.

Code:
if aff_side = 0 my_dv = right.
if aff_side = 1 my_dv = left.
exe.
Then press Ctrl-A to highlight all of the syntax, then Ctrl-R to run it.
 
#5
I would suggest to first take the individual differences. That way you would eliminate the the individual factor and the individual would act as her own control. Just as usual in a paired t-test.

If the data are recorded like in variables like "strength_left" and "strength_right", then I would take the difference (and create the variable "ind_dif"):

ind_dif = strength_left - strength_right

If there is no difference between affected and not affected side, then you would expect numbers around zero, both positive and negative values.

(I guess here that the non-affected side is stronger than the other side on the average)
if the right side is the affected one, then you would expect positive numbers like:

ind_dif = strength_left - strength_right
10 = 21 - 11

if the left side is the affected one, then you would expect negative numbers like:

ind_dif = strength_left - strength_right
-11 = 12 - 23

So then you need to change the sign on "ind_dif" by multiplying by minus one ( -1).

That can be done by transforming the "affected" variable which so far takes values (0=right and 1= left).

affected_new =2*(affected - 0.5)

The affected_new will take value = +1 for left side and -1 for right side.

Now you need to multiply "ind_dif" by affected_new to:

well_vs_affected = ind_dif*affected_new

If you get many positive values, that is the mean on "well_vs_affected" is positive, then there is indication that the "well side" is stronger than the affected side.

But you also had two treatments. Then you can run the variable "well_vs_affected" with a two sample independent t-test. (An equivalent result will be given if you run an one way analysis of variance (anova) with "well_vs_affected" as dependent variable and treatment as an explanatory/independent variable.)