+ Reply to Thread
Results 1 to 2 of 2

Thread: Regression to the mean for discrete non normal data

  1. #1
    Points: 2,673, Level: 31
    Level completed: 49%, Points required for next Level: 77

    Thanked 0 Times in 0 Posts

    Regression to the mean for discrete non normal data


    At work an initiative was started that attempted to reduce the number of Emergency department attendances by frequent users. I have a set of paired data giving the count of A&E attendances for 108 patients covering 12 months pre and 12 months post intervention.

    The distributions for both counts are heavily skewed (positive) and a median test and Wilcoxon test both show a significant reduction in attendances post intervention. I've been asked to account for regression to the mean in the pre intervention counts and retest but am struggling to do so.

    I have read that the amount of regression to the mean can be accounted for by the inverse of the correlation coefficient (1-r). As there are ties in the data I've used the Kendall correlation (tau = 0.29) and modelled the pre-intervention counts with x - (x - xbar) * (1 - tau) resulting in continuous variables which for discrete data can not be.

    Also as the data is heavily skewed I'm wondering if I can use the mean. Would the median be better (regression to the median) or perhaps quartile regression?

    So a little/lot lost and any suggestions would be most welcome.


    Cheers in advance
    Last edited by MrScotchpie; 08-21-2015 at 08:11 AM.

  2. #2
    Omega Contributor
    Points: 38,303, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Not Ames, IA
    Thanked 1,185 Times in 1,146 Posts

    Re: Regression to the mean for discrete non normal data

    How are patients brought to the ED? I am guessing that you could have an issue of patients not always going to your hospital. Can you account for the number of times that they may go to another hospital. If not, you really don't know much here.

    I worked on a project last year on trying not to get certain patients to use the ED unnecessarily (as there primary care), though if I don't know if they are being appropriately admitted to another hospital if we don't admit them, how do I know is I should never admitted them. With your "Cheers" salutation - perhaps your population or patient service use data is better than mine.
    Stop cowardice, ban guns!

+ Reply to Thread


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

Advertise on Talk Stats