Prev: preview video on different axes
Next: Forcing display output from cellarray
From: Roy on 16 Apr 2010 21:24
Hi All,
I would like to get ideas on a data structure for storing multivariate time series
type data.

Background:
I have n experimental conditions with m environmental parameters.
My coworkers are interested in examining how i possible product versions
behave under the above conditions measuring j variables which are all time
series. For statistical purposes, they look at replicates of those multivariate
time series but the kicker is that the number of replicates may vary per time
point.

Possible solution:
I looked at constructing a structure to store the above information.

Below is an example for one experimental condition (EC)
EC(1).NameOfEnvironParameters = vector of environ parameters
EC(1).ValuesOfEnvironParameters = vector of above parameters
EC(1).NameOfJVariables = vector of variable names stored in Rep fields
EC(1).ValuesOfEnvironParameters(1).ProductVersion(1).Name =name of product version
EC(1).ValuesOfEnvironParameters(1).ProductVersion(1).Rep(1) = multivariate time series of one replicate for one product version

EC(1).ValuesOfEnvironParameters(3).ProductVersion(5).Rep(2) = multivariate time series of second replicate for product version 5 under experimental condition3

I've opted for the element-by-element approach to perserve connectivity
between the different fields. But this looks awfully ugly. I was
wondering if there were other options to look at. I've had issues with getting
dataset to work but my feeling is the above is too complicated for even dataset,
but if someone could show me, I'd be happy to take a look at a dataset implementation. Recall the number of replicates for a given time point is not
constant.

Thanks,
RK
 | 
Pages: 1
Prev: preview video on different axes
Next: Forcing display output from cellarray