which test for trend?

[AB365]

Hi
I am afraid I have a pretty basic question, but I just can't seem to find an exact answer that I am happy with. It's the usual thing - which test to use? I have some antimicrobial resistance surveillance data (i.e. X% E. coli resistant to ampicillin) for 7 points of time (the month of February for 7 consecutive years), and I just want to test whether proportion resistance is increasing or not. From my (very rusty) statistics knowledge I would have tried a Chi-squared test for trend, however when I went to go and confirm for myself that this is the correct test to use .. I found myself more confused than ever..
Any help appreciated, apologies if this has been answered already
Aline

[Maguin, Eugene]

Aline,

Maybe somebody else will understand better but I don't quite understand your
data. Do you have seven annual samples where each sample has 10, 100, or
1000 records?

Or do you have one record for each year, like 2001: 18%; 2002: 22%; 2003
17%? Seven records total? I'm assuming you have an N or Ns somewhere because
without that I don't think you can do much because you can't compute the
standard error.

Gene Maguin


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
AB365
Sent: Friday, January 14, 2011 4:02 PM
To: [hidden email]
Subject: which test for trend?

Hi
I am afraid I have a pretty basic question, but I just can't seem to find an
exact answer that I am happy with. It's the usual thing - which test to use?
I have some antimicrobial resistance surveillance data (i.e. X% E. coli
resistant to ampicillin) for 7 points of time (the month of February for 7
consecutive years), and I just want to test whether proportion resistance is
increasing or not. From my (very rusty) statistics knowledge I would have
tried a Chi-squared test for trend, however when I went to go and confirm
for myself that this is the correct test to use .. I found myself more
confused than ever..
Any help appreciated, apologies if this has been answered already
Aline
--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/which-test-for-trend-tp3341921
p3341921.html
Sent from the SPSSX Discussion mailing list archive at 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

[AB365]

Thanks for your reply, I really appreciate it. Sorry I wasn't clear enough. The data I have are repeated period prevalence studies. So for the month of Feb each year I have resistance data on every urine sample which tested positive for bacterial species in a geographic region (first positive sample per patient as far as I could identify). This data is surveillance data, so probably not up to research standards, however I do have data on 8000+ samples in total and this data is not something that is available for this area in published literature. That is why I am interested in trends, because hopefully whatever errors there are are stable over the time period, so even if the resistance rates aren't exactly accurate the trends should show which way things are heading. I don't have the exact data in front of me but it would look something like this for a particular organism-antibiotic combination:

Feb 2004 S=250 R=250 (50%R)
Feb 2005 S=300 R=270 (47%R)
Feb 2006 S=270 R=250 (48%R)
Feb 2007 S=250 R=300 (54%R)
Feb 2008 S=210 R=230 (52%R)
Feb 2009 S=285 R=300 (51%R)
Feb 2010 S=230 R=250 (52%R)

Thanks for your help
Aline

[Maguin, Eugene]

Aline,

Ok, so you have 8000+ records over 7 years. Each record is 'yes/no' with
respect to resistance. It also sounds like each of the 8000 records probably
is a separate individual. Perhaps not completely but you certainly haven't
got 7 years of samples for 1000+ persons. My suggestion is to use logistic
regression. Sounds like Year (0-6) is your only IV. You could treat year as
continuous or categorical. Logistic is correct for a categorical DV.

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
AB365
Sent: Monday, January 17, 2011 3:43 PM
To: [hidden email]
Subject: Re: which test for trend?

Thanks for your reply, I really appreciate it. Sorry I wasn't clear enough.
The data I have are repeated period prevalence studies. So for the month of
Feb each year I have resistance data on every urine sample which tested
positive for bacterial species in a geographic region (first positive sample
per patient as far as I could identify). This data is surveillance data, so
probably not up to research standards, however I do have data on 8000+
samples in total and this data is not something that is available for this
area in published literature. That is why I am interested in trends, because
hopefully whatever errors there are are stable over the time period, so even
if the resistance rates aren't exactly accurate the trends should show which
way things are heading. I don't have the exact data in front of me but it
would look something like this for a particular organism-antibiotic
combination:

Feb 2004 S=250 R=250 (50%R)
Feb 2005 S=300 R=270 (47%R)
Feb 2006 S=270 R=250 (48%R)
Feb 2007 S=250 R=300 (54%R)
Feb 2008 S=210 R=230 (52%R)
Feb 2009 S=285 R=300 (51%R)
Feb 2010 S=230 R=250 (52%R)

Thanks for your help
Aline
--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/which-test-for-trend-tp3341921
p3345084.html
Sent from the SPSSX Discussion mailing list archive at 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]

The data shown below were for one particular organism-antibiotic combination.  So I think organism, antibiotic, and organism*antibiotic are other terms to include in the model.  Various interactions involving Year may be of interest too.  

HTH.

Gene Maguin wrote

Aline,

Ok, so you have 8000+ records over 7 years. Each record is 'yes/no' with
respect to resistance. It also sounds like each of the 8000 records probably
is a separate individual. Perhaps not completely but you certainly haven't
got 7 years of samples for 1000+ persons. My suggestion is to use logistic
regression. Sounds like Year (0-6) is your only IV. You could treat year as
continuous or categorical. Logistic is correct for a categorical DV.

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[e-mail]] On Behalf Of
AB365
Sent: Monday, January 17, 2011 3:43 PM
To: [e-mail]
Subject: Re: which test for trend?

Thanks for your reply, I really appreciate it. Sorry I wasn't clear enough.
The data I have are repeated period prevalence studies. So for the month of
Feb each year I have resistance data on every urine sample which tested
positive for bacterial species in a geographic region (first positive sample
per patient as far as I could identify). This data is surveillance data, so
probably not up to research standards, however I do have data on 8000+
samples in total and this data is not something that is available for this
area in published literature. That is why I am interested in trends, because
hopefully whatever errors there are are stable over the time period, so even
if the resistance rates aren't exactly accurate the trends should show which
way things are heading. I don't have the exact data in front of me but it
would look something like this for a particular organism-antibiotic
combination:

Feb 2004 S=250 R=250 (50%R)
Feb 2005 S=300 R=270 (47%R)
Feb 2006 S=270 R=250 (48%R)
Feb 2007 S=250 R=300 (54%R)
Feb 2008 S=210 R=230 (52%R)
Feb 2009 S=285 R=300 (51%R)
Feb 2010 S=230 R=250 (52%R)

Thanks for your help
Aline
--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/which-test-for-trend-tp3341921
p3345084.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (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
[e-mail] (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

[AB365]

Thanks for your help. Much appreciated.
Aline