+ Reply to Thread
Results 1 to 7 of 7

Thread: Random Effects Regression vs. Fixed Effects Regression

  1. #1
    Points: 4,647, Level: 43
    Level completed: 49%, Points required for next Level: 103

    Posts
    19
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Random Effects Regression vs. Fixed Effects Regression




    I'm having trouble distinguishing when to use random effects versus fixed effects regression. Any thoughts?

  2. #2
    Super Moderator
    Points: 19,965, Level: 89
    Level completed: 23%, Points required for next Level: 385
    bugman's Avatar
    Posts
    1,884
    Thanks
    200
    Thanked 231 Times in 185 Posts
    Random effects are generally used if the levels of the independent variable are thought to be a small subset of all possible values. For example, if were looking at growth rates of a plant on five randomly selected types of soil of all the soil types. In this case you may be just wanting to know if growth rates differed depending on soil.

    With fixed effects, you are fixing your independant variable: so that you are specifically interested in certain soils. In this case you chose, or set the types of soil. Now your question might become more specific relating to specific soils.

    This is pretty simplistic but I hope it helps.

    Was are the variables you are looking at?

  3. #3
    Points: 4,647, Level: 43
    Level completed: 49%, Points required for next Level: 103

    Posts
    19
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Simple examples are always welcomed.

    I'm studying multiple hospitals, but have patient data on hip surgery. My independent variable is focused care (continuous variable) and the dependent variable is day till patient can get out of bed post-surgery.

    If I understand you correctly, then I want to know how different types of focused care that the different hospitals deliver affects the patient's mobility after hip surgery.

    I don't think I understand what you mean by this: "the levels of the independent variable are thought to be a small subset of all possible values"

  4. #4
    Super Moderator
    Points: 19,965, Level: 89
    Level completed: 23%, Points required for next Level: 385
    bugman's Avatar
    Posts
    1,884
    Thanks
    200
    Thanked 231 Times in 185 Posts
    RE: the levels of the independent variable are thought to be a small subset of all possible values

    So, if you want to know say, if survival (or hip care) was dependant on the hospital they were treated at; you might not have the time, money or resources to sample all the hospitals in your city, state, country or whatever so you draw a subset of these from all the possible hospitals (i.e. the popualtion).
    This is a case of treating it (hospital) as a random effect.

    On the other hand, if you specifcally wanted to know if Hospital A was different from hospital B and so on - this would be an example of a fixed effect (i.e. you only care about these and none of the others in the popualtion).
    Hope this clarifies it a little.

    CAn you clarify focused care? How is it measured?

  5. #5
    Points: 4,647, Level: 43
    Level completed: 49%, Points required for next Level: 103

    Posts
    19
    Thanks
    0
    Thanked 0 Times in 0 Posts
    Focused care is a composite measure that we created on a scale of 1 to 10. The indicator represents the extent to which the hospital dedicates staff time, facility space, $$, technology, and protocols for hip surgery. The more resources focused on hip surgery, the more specialized the hospital. ...And you would hypothesize, that this would also mean better mobility outcomes for the patient because the care is super specialized.

    I understand your explanation perfectly, but now I'm uncertain if I should be using random effects or fixed effects. I have 21 hospitals, supposedly a subset of the universe of hospitals, but I do want to tease out the effect of the different degrees of focus in these hospitals on the outcome. I don't necessarily want to know the effect of focus for these 21 hospitals, but for the universe of hospitals with degrees 1 through 10.

    Now that I've said that, it is random effects that I should do, isn't it?

  6. #6
    Super Moderator
    Points: 19,965, Level: 89
    Level completed: 23%, Points required for next Level: 385
    bugman's Avatar
    Posts
    1,884
    Thanks
    200
    Thanked 231 Times in 185 Posts
    Treat Hospital as a random effect if its in your model and focused care as a fixed effect since it has boundaries and you are specifically interested in the various levels.

  7. #7
    Points: 4,647, Level: 43
    Level completed: 49%, Points required for next Level: 103

    Posts
    19
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Ah, I see. Thanks for your help Bugman!

+ Reply to Thread

           




Similar Threads

  1. Panel Data - Pooled OLS vs Fixed Effects vs Random Effects
    By g1362901 in forum Statistical Research
    Replies: 2
    Last Post: 03-09-2012, 12:54 PM
  2. doing a fixed effects regression
    By lm8ub in forum Stata
    Replies: 1
    Last Post: 02-25-2011, 03:19 AM
  3. Fixed Effects Common Panel Regression versus GMM estimation
    By Herbert in forum Regression Analysis
    Replies: 0
    Last Post: 02-18-2011, 06:43 AM
  4. Quick Q on Fixed vs. Random effects
    By brownmoses in forum Statistics
    Replies: 0
    Last Post: 10-08-2009, 01:17 PM
  5. Time demeaned v. LSDV fixed effects regression
    By Jared in forum Statistics
    Replies: 0
    Last Post: 11-03-2007, 08:28 AM

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