0
$\begingroup$

I have a general knowledge of regression, but I don't know what is the best approach to analyze these data. The dataset represents quarterly rates of inpatient use for three provider groups prov1, prov2 and prov3. The rates were measured on quarterly basis from qrt-4 2005 until qrt-3 2012. On the 3rd qrt of 2008 an intervention was implemented by provider groups 1 and 2, provider 3 represents all other providers of a network that did not participate of the intervention; so this is our control group. The purpose of this project is to measure the effect of the intervention between the before and after periods and between the participating and non-participating providers. Here is an example of the data:

Qrt-Yr  Period  Prov1   prov2   Prov3
Q4-2005 Before  103.4   201.1   189.9
Q1-2006 Before  175.4   224.2   184.5
Q2-2006 Before  140.2   237.3   187.7
Q3-2006 Before  196.8   255.6   183.5
Q4-2006 Before  222.1   204.7   187.7
Q1-2007 Before  172.5   212.5   189.7
Q2-2007 Before  172.7   207.9   185.9
Q3-2007 Before  128.8   214.7   192.9
Q4-2007 Before  164.9   197.7   205.5
Q1-2008 Before  172.1   225.9   210.3
Q2-2008 Before  103.9   209.3   203.1
Q3-2008 Before  151.1   223.7   211.4
Q4-2008 After   200.8   234.3   189.6
Q1-2009 After   158.1   185.6   206.8
Q2-2009 After   103.8   218.2   187.9
Q3-2009 After   132.7   204.6   202.5
Q4-2009 After   178.7   218.7   191.3
Q1-2010 After   181.0   223.6   178.4
Q2-2010 After   119.9   203.2   191.3
Q3-2010 After   144.6   228.7   193.8
Q4-2010 After   87.2    161.0   173.1
Q1-2011 After   102.4   178.7   175.8
Q2-2011 After   151.0   173.4   169.9
Q3-2011 After   137.1   233.9   187.5
Q4-2011 After   83.0    226.6   164.3
Q1-2012 After   147.6   193.7   172.0
Q2-2012 After   103.8   163.5   171.7
Q3-2012 After   159.1   180.8   172.5
Q4-2012 After   133.1   155.1   160.7
$\endgroup$
1
$\begingroup$

I do recommend you reconstruct your data as "retail month" data instead of quarterly so you will have enough time points for a good analysis. A retail month is a constant 28 days long and controls for the fact that the number of weekend days varies across months introducing a systematic component that would need to be modeled.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.