| Author |
Message |
W. D. Allen
science forum beginner
Joined: 27 Jun 2006
Posts: 1
|
 Posted: Tue Jun 27, 2006 10:45 pm Post subject:
Correlation Of Hospital Admittances
|
|
|
Need an approach that allows me to analyze the following data. I have five
sets of monthly admittance counts versus total insurance enrollment data for
each of five hospitals. I need to determine if any of the hospitals are
performing subpar, that is, too many admittances for the corresponding
enrollments compared to the other four hospitals.
My approach is to sum the hospital percentage data by month for all five
hospitals to construct a control group. Then I will calculate the
correlation of each of the five against that control group.
Question - is there a more statistically elegant approach to identifying any
subpar performers in the group?
WDA
ballensr@adelphia.net
end |
|
| Back to top |
|
 |
dave@autobox.com
science forum beginner
Joined: 16 Feb 2006
Posts: 12
|
| Posted: Wed Jun 28, 2006 11:59 am Post subject:
Re: Correlation Of Hospital Admittances
|
|
|
W. D. Allen wrote:
| Quote: | Need an approach that allows me to analyze the following data. I have five
sets of monthly admittance counts versus total insurance enrollment data for
each of five hospitals. I need to determine if any of the hospitals are
performing subpar, that is, too many admittances for the corresponding
enrollments compared to the other four hospitals.
My approach is to sum the hospital percentage data by month for all five
hospitals to construct a control group. Then I will calculate the
correlation of each of the five against that control group.
Question - is there a more statistically elegant approach to identifying any
subpar performers in the group?
WDA
ballensr@adelphia.net
end
|
B.
Your suggested approach leaves a lot to be desired IMHO.
To begin with correlation is meaningless when you have time series
data...
If I had this task I would identify and build 5 models relating monthly
admittance to enrollment data making sure that any neeede lag structure
in enrollments was incorporated and any necessary ARIMA structure for
unspecified stochastic series was also incorporated aling with ant
Pulses, Level Shifts, Seasonal Pulses ( like a December effect ) and.or
any Local Time Trends were incorporated.
I would then be able to specify a GLOBAL MODEL which I would then
estimate using ALL 5 SETS of data and then form an F test to test the
hypothesis of a COMMON SET OF PARAMETERS over the 5 groups. This is
sometimes referred to as pooled cross-sectional time series analysis or
panel data analysis.
I would then be able to idebtify which ( if any !) of the 5 had
statistically significant parameters from EACH OTHER much like a
Scheffe type test for determining which means are different from each
other.
For more on time series see http://www.autobox.com .
If you wish to chat further .... please feel free to call me .
Dave Reilly
Automatic Forecasting Systems
215-675-0652 |
|
| Back to top |
|
|
Nigel
science forum beginner
Joined: 03 Jun 2005
Posts: 37
|
| Posted: Wed Jun 28, 2006 2:51 pm Post subject:
Re: Correlation Of Hospital Admittances
|
|
|
W. D. Allen wrote:
| Quote: | Need an approach that allows me to analyze the following data. I have five
sets of monthly admittance counts versus total insurance enrollment data for
each of five hospitals. I need to determine if any of the hospitals are
performing subpar, that is, too many admittances for the corresponding
enrollments compared to the other four hospitals.
My approach is to sum the hospital percentage data by month for all five
hospitals to construct a control group. Then I will calculate the
correlation of each of the five against that control group.
Question - is there a more statistically elegant approach to identifying any
subpar performers in the group?
WDA
ballensr@adelphia.net
end
|
I don't fully understand what you're doing, but wouldn't it be better to
compare each hospital in turn against the sum of the other four?
NigelH |
|
| Back to top |
|
|
dave@autobox.com
science forum beginner
Joined: 16 Feb 2006
Posts: 12
|
| Posted: Wed Jun 28, 2006 9:21 pm Post subject:
Re: Correlation Of Hospital Admittances
|
|
|
nigel wrote:
| Quote: | W. D. Allen wrote:
Need an approach that allows me to analyze the following data. I have five
sets of monthly admittance counts versus total insurance enrollment data for
each of five hospitals. I need to determine if any of the hospitals are
performing subpar, that is, too many admittances for the corresponding
enrollments compared to the other four hospitals.
My approach is to sum the hospital percentage data by month for all five
hospitals to construct a control group. Then I will calculate the
correlation of each of the five against that control group.
Question - is there a more statistically elegant approach to identifying any
subpar performers in the group?
WDA
ballensr@adelphia.net
end
I don't fully understand what you're doing, but wouldn't it be better to
compare each hospital in turn against the sum of the other four?
NigelH
|
Not really since any one non-representative data point ( anomaly )
would distort the correlation coefficient. Since the values used to
compute the correlation coeeficient are autocorrelated the validity of
the bivariate normality assumption is immediately in question.
Time series data produces spurious correlation coefficients particulary
when one of the series is non-stationary ( like trends or level shifts
or strong ar structure )
dave r
afs |
|
| Back to top |
|
|
Google
|
|
| Back to top |
|
|
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
Forum index
»
Science and Technology
»
Math
»
Probability
»
Prediction
