chi square or unpaired t-test? (or other??)

#1
Hi -
I am looking to compare two groups - mobility activity after surgery.

Group 1 - is a 'low ability' (ie, very old, frail)
Group 2 - is 'high ability' (able to move around before surgery).

The numbers in each group are NOT means, they are simply counts of patients. Ie, on the first day, 37 of Group 1 got up and moved around, and 25 of Group 2.
I have numbers for 5 days. Not continuous (ie, they are not the same patients each day - just counts of patients moving in each group).
I would like to compare the 2 groups over the 5 days to see if there is a significant difference between each group.

ANOVA - is for comparing means - which is not applicable here? (I think).

Initially I was thinking chi square, but now I am wondering if a t-test would be best?
thank you for your comments!
 

Karabiner

TS Contributor
#2
I have numbers for 5 days. Not continuous (ie, they are not the same patients each day - just counts of patients moving in each group).
Does that mean, you have 2*5=10 completely distinct groups, no patient e.g. from day 1 can appear in days 2 to 5?

With kind regards

Karabiner
 
#3
Yes and no.
I would like to keep them as two groups, (low and high), across 5 post operative days.
Across all days - is there a difference between the two groups. The question you ask may be the problem that I can't get it 'right' in SPSS.

I simply listed the two groups, each with 5 sets of data.

But you are also correct, the numbers are counts, a patient may represent more than one 'count' through the 5 days (there were 77 patients in total), but I have not entered my data as patients, rather than total count of patients (does this make sense?) POD=post operative day

________ Low ____ High
Pod 1 _____ 34 ____ 37
Pod 2 _____ 15 ____ 31
Pod 3 _____ 16 ____ 21
Pod 4 _____ 8 ____ 15
Pod 5 _____ 9 ____ 12

Would you recommend entering as individual patients? I am not sure this would make a difference...
 
Last edited:

Karabiner

TS Contributor
#4
If the same patients are measured across 5 days, this means you
have dependent measures across days, and you should preferably
collect individual responses. Since you only have aggregated responses,
you can perform 5 separate Chi² analyses (group [Low/High] vs. moving
[yes/no]), one for each day. You could introduce some kind of correction
for increased risk of false-positive results, e.g. Bonferroni (here,
alpha 1% instead of the usual 5%).

With kind regards

Karabiner
 
#5
If the same patients are measured across 5 days, this means you
have dependent measures across days, and you should preferably
collect individual responses. Since you only have aggregated responses,
you can perform 5 separate Chi² analyses (group [Low/High] vs. moving
[yes/no]), one for each day. You could introduce some kind of correction
for increased risk of false-positive results, e.g. Bonferroni (here,
alpha 1% instead of the usual 5%).

With kind regards

Karabiner

thanks so much for your feedback~~!