+ Reply to Thread
Results 1 to 5 of 5

Thread: Binary Logistic Regression and Generalised Linear Model

  1. #1
    Points: 983, Level: 16
    Level completed: 83%, Points required for next Level: 17

    Posts
    29
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Binary Logistic Regression and Generalised Linear Model




    I have basic understanding of statistics and am trying to interpret the output of a generalised linear model.

    The data used in my analyses pertains to factors associated with visual acuity gain (indicated by a decrease of 0.1 LogMAR).

    I have two outcomes for visual acuity gain.
    i) 1 categorical (No Gain, Gain)
    ii) 1 scale (e.g. 0.3 LogMAR)

    After reading some online texts, I found that a binary logistic regression could be used for the categorical variable while a generalised linear model is suited to the scale variable.

    At present I have produced a binary logistic regression which I can understand well, however the generalised linear model leaves me in a state of confusion.

    The factors included as main effects include gender (male and female), visual acuity at baseline (scale), astigmatism (scale), intraocular power (scale), axial length (scale) and age (scale).

    From my understanding I gather that the generalised linear model is more appropriate as there is finer gradation for the scale variable as opposed to Gain versus No Gain.

    1) What is meant by “set to zero because this parameter is redundant” and “maximum likelihood estimate”?

    2) Could someone recommend a published article that presents results as a GLM?

    I appreciate any comments or advice.

    Best,

    J
    Last edited by jle; 01-22-2015 at 12:15 AM.

  2. #2
    Fortran must die
    Points: 58,790, Level: 100
    Level completed: 0%, Points required for next Level: 0
    noetsi's Avatar
    Posts
    6,532
    Thanks
    692
    Thanked 915 Times in 874 Posts

    Re: Binary Logistic Regression and Generalised Linear Model

    I can't help with a generalized linear model, but I have some basic comments. If you have basic understanding of statistics you should stick with the simplest method usable. For one thing there are nuances, particularly in assumptions of advanced methods that you may miss with serious consequences. And less advanced models are generally better covered in the literature. And they may well work well enough for your analysis. I have worked with regression for years now and still try to use the simplest method I can for this reason.

    What is the dependent variable that you feel you need to use general linear models for? That is, is it an ordinal or nominal variable and how many distinct levels does it have? Why is linear or logistic regression inappropriate for it?
    "Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995

  3. #3
    Points: 983, Level: 16
    Level completed: 83%, Points required for next Level: 17

    Posts
    29
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Binary Logistic Regression and Generalised Linear Model

    Thank you Noetsi!

    While I agree that is is useful to analyse data using the simplest method, I am very adamant to understand the GLM output I have generated.

    The dependent variable is a scale variable (continuous) for visual acuity gain. Both are appropriate, however there are some things that remain unclear to me as listed above.

  4. #4
    Fortran must die
    Points: 58,790, Level: 100
    Level completed: 0%, Points required for next Level: 0
    noetsi's Avatar
    Posts
    6,532
    Thanks
    692
    Thanked 915 Times in 874 Posts

    Re: Binary Logistic Regression and Generalised Linear Model

    I don't know your field at all, but is this related enough?

    The aim of this study was to determine whether typical abnormalities seen on autofluorescence (AF) imaging in patients with macular telangiectasia (MacTel) type 2 are correlated with visual acuity at presentation and with progression of visual loss over a 2-year follow-up period.
    The article used generalized linear models. I ask because generally it is most helpful to find something in the literature related to what you are doing that uses the method. Even if the results you want are not in the article the author might be willing to assist you.


    http://www.ncbi.nlm.nih.gov/pubmed/24743635

    Think this is the same article. If you are at a university you can probably get access to it.

    http://onlinelibrary.wiley.com/doi/1...466.x/abstract

    I think the answer to this question:

    1) What is meant by “set to zero because this parameter is redundant” and “maximum likelihood estimate”?
    Will have to be found in the specific software you are using (that is somewhere in the documentation or help desk). Software uses very different comments in my experience for the same statistical reality. Not sure what software you used.
    "Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995

  5. The Following User Says Thank You to noetsi For This Useful Post:

    kiton (01-25-2015)

  6. #5
    Points: 983, Level: 16
    Level completed: 83%, Points required for next Level: 17

    Posts
    29
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Binary Logistic Regression and Generalised Linear Model


    These are a bit more advanced than my current project, although I really appreciate your comments. I think have made some progress though which I'm happy with.

    Thanks again Noetsi!

+ 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