comparing proportions

[LiZ]

Can anybody help me finding the right statistical test to use? I would like
to compare the difference in proportion of resolved disease variables
between two treatment groups. The patients were followed over time, at three
different time points (week 2, 4 and 18). There are three different disease
variables. Those disease variables are scored based on acquired images. The
scores were either resolved or not resolved. So I would like to compare the
proportion of resolved scores between the groups, for each disease variable
and follow-up moment.

Find attached an example of my data.


forum_vraag.xlsx
<http://spssx-discussion.1045642.n5.nabble.com/file/t341575/forum_vraag.xlsx>  



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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

[Rich Ulrich]

I've got problems with the problem. I thought that "resolved" means

that the disease state (rash, whatever) goes away.  That is one sort of

problem, where longer followup does not matter.  Once it is gone, you

have a different problem if it is possible for it to COME BACK.  Then -

there seems to be missing data, since N declines for weeks 4 and 18.

A second problem is "lack of power".  For variable A, I see an "effect"

that is decent sized at week 2:  3 resolved of 13, vs 5 resolved of 9,

giving an Odds ratio of almost 2 while counts are greater than 1.

But this is not at all statistically significant.   Other single comparisons

for A and B have (I think) even less power.

In my experience, I would want to know if treatment X is better than Y

as my main question and main test.  Since not everyone has every symptom

(as I interpret the differing Ns at 2), what is countable seems to be the

number of "unresolved symptoms" for each person.  The simplest test at

week 2 would be a t-test between counts for groups X and Y.  If the count

of symptoms is not equal at week 0 (not shown), the test controlling for

initial status would be the one-way Ancova, where Unresolved-week 0 is

the covariate while comparing the Unresolved-week 2.

Dealing fairly with later weeks is confused by the lack of information

about those Missing followups.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of LiZ <[hidden email]>
Sent: Friday, February 15, 2019 5:04 AM
To: [hidden email]
Subject: comparing proportions

 

Can anybody help me finding the right statistical test to use? I would like
to compare the difference in proportion of resolved disease variables
between two treatment groups. The patients were followed over time, at three
different time points (week 2, 4 and 18). There are three different disease
variables. Those disease variables are scored based on acquired images. The
scores were either resolved or not resolved. So I would like to compare the
proportion of resolved scores between the groups, for each disease variable
and follow-up moment.

Find attached an example of my data.


forum_vraag.xlsx
<http://spssx-discussion.1045642.n5.nabble.com/file/t341575/forum_vraag.xlsx> 



--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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

===================== 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

[Bruce Weaver]

Like Rich, I find the layout of your data file confusing.  From what you've
said about your data, I expected a LONG file format with the following
variables:

  ID - unique identifier for each individual
  Disease - disease code
  Week
  Resolved - 1=Yes, 0=No

This is the layout required to estimate a multilevel logistic regression
model via GENLINMIXED (with week clustered within ID), for example.  This
book by Heck, Thomas & Tabata is a good resource, if that is indeed the type
of model you want:

  http://www2.hawaii.edu/~ltabata/mlm/HTT2012.html

Alternatively, you could use GENLIN with generalized estimating equations
(GEE) to account for the correlated nature of the repeated measures (over
weeks).  

Another thing that's not clear to me is whether you want to estimate
separate models for each disease, or if you want disease to be a factor in
one overall model.  

HTH.



LiZ wrote

> Can anybody help me finding the right statistical test to use? I would
> like
> to compare the difference in proportion of resolved disease variables
> between two treatment groups. The patients were followed over time, at
> three
> different time points (week 2, 4 and 18). There are three different
> disease
> variables. Those disease variables are scored based on acquired images.
> The
> scores were either resolved or not resolved. So I would like to compare
> the
> proportion of resolved scores between the groups, for each disease
> variable
> and follow-up moment.
>
> Find attached an example of my data.
>
>
> forum_vraag.xlsx
> &lt;http://spssx-discussion.1045642.n5.nabble.com/file/t341575/forum_vraag.xlsx&gt; 
>
>
>
> --
> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (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





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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