Please help me in the following statistical models

[Grisha]

Dear Users,

I'm looking for some help to answer 4 statistical questions in a complex research.
I'm quite new in SPSS and any help would be appreciated.


Please help me find the suitable statistical tests to answer the following question.

1.

There were 3 types of treatment and 5 measured parameter,

The columns contains the given measured parameters, the rows are all the measured parameters for a subject.

I have a column that contains 1,2 or 3 (1: treatment A, 2: treatment B, 3: untreated)

I have 33 columns that contains the 5 measured parameters in different timepoints (for eg. parameter A 1hour, parameter A 6 hours, parameter A 1 month ..., etc )


The question: Does the chosen treatment affect the measured parameters at any timepoint?

2.

Multiply measurement: 5 timepoint, 2 types of measurements, and one global factor

Type 1 measurements: 5 measured parameter in 5 timepoints
Type 2 measurements: 12 measured parameters in the same 5 timepoint.
Global factor can range from 1 to 8

So every patient (row) has 5x5 values for Type 1 measurements, 5x12 values for Type 2 measurements and one value for global factor.

The questions:

A, Is there any relationship between the the measured Type 1 measurements at the same timepoint (eg. does one type 1 parameter determine another type 1 parameter at the same time)

B, Are the Type 1 measurement parameters affected by the Type 2 measurement parameters or by the global factor?


3.

We have the same Type 1 measurements: 5 measured parameter in 5 timepoints
One column for cardiac function (range 0-100)
And many other binary columns (risk factors, previous illness, etc, as 0 stands for no, 1 for yes)


Question: Are the Type 1 measurements in a single timepoint affected by the cardiac function or by the risk factors.
So I'd need 5 evaluation, 1 for each timepoint.


4.
Are the Type 1 measurement parameters determined/affected by another column (two types of conditions values: 1 or 2)




I'd be really greatful for any help. So far, I only used simple correlation tests in another statistical program, I'm
really new to SPSS.


Thank you very much.



Gergo



ps.: I'm using spss19 on a 64bit win 7

[Maguin, Eugene]

Gergo,

Your explanation of your data leaves me with questions.

>>1. There were 3 types of treatment and 5 measured parameter, The columns
contains the given measured parameters, the rows are all the measured
parameters for a subject. I have a column that contains 1,2 or 3 (1:
treatment A, 2: treatment B, 3: untreated). I have 33 columns that contains
the 5 measured parameters in different timepoints (for eg. parameter A
1hour, parameter A 6 hours, parameter A 1
month ..., etc ). The question: Does the chosen treatment affect the
measured parameters at
any timepoint?

>>2. Multiply measurement: 5 timepoint, 2 types of measurements, and one
global
Factor. Type 1 measurements: 5 measured parameter in 5 timepoints. Type 2
measurements: 12 measured parameters in the same 5 timepoint. Global factor
can range from 1 to 8. So every patient (row) has 5x5 values for Type 1
measurements, 5x12 values for Type 2 measurements and one value for global
factor.

Let's first get the data structure clearly understood. I'm assuming that you
are have the following data structure.

ID Tx Typ1A1 Typ1B1 ... Typ1D5 Typ1E5 Typ2A1 ... Typ2L1 ... Typ2K5 Typ2L5
GF1 ... GF5

Where Typ1A1=a Type 1 variable named A measured at time 1.
GF1=Global factor measured at time 1.


>>The questions:
>>A, Is there any relationship between the the measured Type 1 measurements
at
the same timepoint (eg. does one type 1 parameter determine another type 1
parameter at the same time)

>>B, Are the Type 1 measurement parameters affected by the Type 2
measurement
parameters or by the global factor?

>>3. We have the same Type 1 measurements: 5 measured parameter in 5
timepoints. One column for cardiac function (range 0-100). And many other
binary columns (risk factors, previous illness, etc, as 0 stands for no, 1
for yes). Question: Are the Type 1 measurements in a single timepoint
affected by the cardiac function or by the risk factors. So I'd need 5
evaluation, 1 for each timepoint.

>>4. Are the Type 1 measurement parameters determined/affected by another
column (two types of conditions values: 1 or 2).

You (or somebody) have given yourself a huge job. I think the absolutely
most useful thing for you to do is find somebody familiar with spss and
data/statistical analysis where you are located who can work with you. You
can learn a huge amount from your project; you just need to make sure you
survive it.

That said.
Under the assumption that all measures are continuous enough to be treated
as continuous, that the  treatment group sample sizes are large enough and
minimal missing data due to any cause.
Questions A & B.
'determine' and 'affected by' is incorrect language that implies causal
relationships. Think of 'covary' and 'correlate' instead.
As you have done, begin with correlations then move to regression to control
for treatment group membership. I ignore demographics/other collected
variables; you may not want to. Lastly, consider whether within treatment
group correlation differ across groups by adding a group by independent
variable interaction term. You have the syntax reference file for spss, read
the Regression command documentation.

Gene Maguin

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD