Need help in analysis

#1
Hi....All stats gurus ....kindly help me with this

I have data from 12 patients, and we have measured 5 hemodynamic variables ( Systolic Blood Pressure, Diastoic Blood P, cardiac index, venous resistance and artery complaince) on each of these 12 patients on 4 different days. These measurements were done using a instrument that we made ( which is validated). I also have the data on the weight gain of these patients for these 4 days.

I have subdivided the data based on delta weight gain. i.e <2 lbs <1 and <0.5 ( this is done like this... say patient 1 has 4 weight gains: 1 lb, 3.2 lb, 2.8, 1.7 then we calculate (delta weight gain) highest-lowest (3.2-1=2.2) is this patients delta weight gain. So in order to have this patient in <2 group we will delete 1 ( first value), so tht now delta weight gain is 3.2-1.7=1.5. When we do this we have to delete all 5 hemodynamic variables for that patient for 1st day in order to make sense. To have this patient in <0.5 group- we will delete both 1 and 1.7 to make it 3.2-2.8=0.4. But we remove corresponding values for same patient from all 5 variables.

This generates data for 12 patients with variable # values may be 1,2,3 or 4 ( out of 4 days)for each 5 variables.
This way all 12 patients have been readjusted for delta weight gain <2, <1, <0.5

Now My research question is:
1) are the readings for each hemodynamic variables reproducible for every patient- i mean within patient not between patient, assuming evrything else is stable.

2) is there a relation between delta weight gain and variation in 5 hemodynamic variables i.e lets say when weight gain is stable ( <0.5 variation) then there is less variation of hemodynamic parameters

I was trying to use intra class correlation coefficient and repeated measures ANOVA. but the problem is when subdividing the data based on delta weight gain i have lots of missing values for all of patients. I dont know how to handle that.

I don't know if I was able to explain my data properly.

If anyone can help me with this problem I would greatly appreciate it
 
#3
This is just an idea:

You say that you have subdivided your data based on delta weight gain and that this has caused a lot of missing values, is this correct?

Well, why not just calculate the change in weight for the four time points but not subdivide.

This is what I imagine your design is looking like:


v1 v2 v3 v4 v5 weight change_in_weight1 change_in_weight2 ...
1 .02 .1
2 -.1 -.5
3 .2 -.2
4
5
6
7
8
9
10
11
12

The v1 through v5 are the five variables that you mention, I simplified things because you mentioned also that you have these on four different days. so 'change_in_weight1' means change in weight on day 1, etc.
I filled in some data just for illustration. So if you just leave in the change in weight you are not losing any data. Even if you did want to subdivide your data so there are less categories, you should not be losing any data. You are just recoding your data so that changes in weight that are say between 2 and 10 points would get coded together as 'more than 2 pounds'.

If you are interested in the relationship between say v1 and change_in_weight1, then you could do a correlation analysis. Maybe I'm simplifying things but you could see which of your 5 variables correlate with change_in_weight.

If you don't subdivide your data you will have less missing values and perhaps you could try to do your repeated measures anova.

I hope I understood your problem, if not, feel free to explain it again.
 
E

elnaz

Guest
#4
hello
yes i am agree with reverent CBABD, if its possible change your procedure.
if possible you tell me what is your aim in this project that you did this procedure?
then maybe i could help you in change of procedure.
Best regards
Elnaz