Welcome to the Podiatry Arena forums

You are currently viewing our podiatry forum as a guest which gives you limited access to view all podiatry discussions and access our other features. By joining our free global community of Podiatrists and other interested foot health care professionals you will have access to post podiatry topics (answer and ask questions), communicate privately with other members, upload content, view attachments, receive a weekly email update of new discussions, access other special features. Registered users do not get displayed the advertisements in posted messages. Registration is fast, simple and absolutely free so please, join our global Podiatry community today!

  1. Have you considered the Clinical Biomechanics Boot Camp Online, for taking it to the next level? See here for more.
    Dismiss Notice
Dismiss Notice
Have you considered the Clinical Biomechanics Boot Camp Online, for taking it to the next level? See here for more.
Dismiss Notice
Have you liked us on Facebook to get our updates? Please do. Click here for our Facebook page.
Dismiss Notice
Do you get the weekly newsletter that Podiatry Arena sends out to update everybody? If not, click here to organise this.

Stepwise logistic regression analysis

Discussion in 'General Issues and Discussion Forum' started by Dido, Mar 20, 2008.

Tags:
  1. Dido

    Dido Active Member


    Members do not see these Ads. Sign Up.
    Can someone who know more about statistics/research than I do, explain how the above principle works? I undestand it is a reseach tools for processing data.
    Many thanks
    Dido
     
    Last edited: Mar 20, 2008
  2. Craig Payne

    Craig Payne Moderator

    Articles:
    8
    I used it in: Payne CB, Turner DE, Miller K: Determinants of Plantar Pressure in the Diabetic Foot; Journal of Diabetes and its Complications 16(4)277-283 2002

    A regression model is an equation that uses explanatory variables to predict.
    eg lets assume that:
    plantar pressure under met head one = (2 x declination of first met head) + (3.2 x bodyweight) + (4.2 x Neuropathy 'score') - (8 x thickness of plantar fat pad) + etc etc

    A stepwise logistic regression model is the regression model with the least number of explanatory variables that best explains the plantar pressure under met head one.

    What the software will do is run the regresion equation with just one explanatory variable (eg declination of first met) and calculate the r squared (which is a measure of how well a predictor relates to the plantar pressure). The software then runs again with 2 variables, then 3 variables, then 4 --- this continues until all the variables are used - at each step the r squared is calculated. At the end of this process a combination of some of the variables will be the best set to predict or explain the plantar pressures under met head one. ....

    The above eg is forward stepwise regression -- ie it moves forward by adding variables. The software can also do a backward stepwise regression - its starts with a regression equation with all variables in and then moves backward removing one, then two, then three, etc at a time until the 'best' set of predictors is found.

    Make sense?

    In the above paper we found that:
    The coefficient for the first MPJ ROM was -1.5 and for the Neuropathy score it was 5.5. The -1.5 and 5.5 are like the "slope of the line" in a simple regression equation. The r squared was 0.28 which was the best we could get with the large number of variables that were measured.

    Here is the wikipedia link on the topic.
     
    Last edited: Mar 20, 2008
  3. Dido

    Dido Active Member

    Thank you Craig for that information.
    I will print it out and study it.
    I was reading a research paper concerning the factors responsible for amputations in diabetics. The conclusion of the study was that "using stepwise logistic regression analysis the only factors of any significance were PVD and infection."
    Dido
     
  4. And what was the r square value for the model? The r square tells you what proportion of the variation in the dependent variable, i.e. amputation is "accounted for" by the independents, i.e. PVD and infection.

    Of note here is that amputation is a binary variable, that is you either have amputation or not, did the authors use a binary logistic regression? This is slightly different but wikipedia seems to cover this too.
     
    Last edited: Mar 20, 2008
  5. Dido

    Dido Active Member

    Hello Simon,
    I do not know what the r square was for the model. :confused:
    The paper concerned was:-
    Epidemiology of Diabetic Foot Problems and Predictive Risk Factors for Limb Loss.
    Aziz Nather, Chionh Siok Bee, Chan Yiong Huak, Jocelyn L L Chew, Clarabelle B Lin, & Shuhui Neo.
    Journal of Diabetes and it's Complications.
    Vol 22, Issue 2, March-April 2008, pages 77-82
    Dido
     
  6. You need to know this, was it not reported?
     
  7. Dido

    Dido Active Member

    Hello Simon,
    I think I am way out of my depth looking at these kinds of statistical analyses.
    All I have is "The predictive factors for limb loss were determined using univariate and stepwise logistic regression analysis."
    I don't want to say much more about the paper in case it is copyright.
    Dido
     
  8. I'm fairly sure you can report the r square value here without fear of copyright infringement.
     
  9. Dido

    Dido Active Member

    Hello Simon,
    I don't see it listed anywhere.
    Why is it important?
    Dido
     
  10. Craig Payne

    Craig Payne Moderator

    Articles:
    8
    Without looking at the paper, what I assume they would have done is collected a whole lot of data on the variables that might explain an amputation.

    They then could have (and many papers do this) present the data in a univariate form ... ie a table of the relationship of each (ie 'univariate') variable to the amputation .... its not really this, but sort of consider this as a correlation between that variable and amputation.

    Multivariate analysis is much more robust when there are 'multi' variables - multiple regression is just one type of multivariate analysis.

    In the paper they would have put the data on all the potential explanatory variables for the amputation into the software. It would have found the best subset or combination of those variables that best predict the amputation --- in this case, all that came out were infection and PVD (other variables the software may have discarded could have been age, duration of diabetes, lipid levels, foot structure, etc etc).

    The r squared is a statistic associated with how well the model fits (its not strictly that, but is a good idea to think it of that if you unfamiliar with stats) - the higher the r squared, the better the model (which is why Simon asked for it).
     
  11. Dido

    Dido Active Member

    Hello Craig,
    You have summed up the paper pretty well - I am impressed! I was looking at this paper as background reading for my CPD in UK where diabetes is a core subject.
    Thank you for explaining this to me.
    I am not familiar with statistics or processing data in this way and think I will avoid it in future as it all seems very complex. I am a generalist podiatrist and not a researcher, I think I have wandered out of my depth.
    But again, thanks for your help.
    Dido
     
Loading...

Share This Page