+ Reply to Thread
Results 1 to 3 of 3

Thread: PhD Dissertation Help- General guidance and direction desperately needed!!!!

  1. #1
    Points: 10, Level: 1
    Level completed: 19%, Points required for next Level: 40

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    PhD Dissertation Help- General guidance and direction desperately needed!!!!




    Hi,

    I have been reading many of your comments and I was hoping that you could give me advice. I'm currently in my second year of the dissertation process of my PhD. My committee chair has not been very helpful in guiding me in which type of statistical analysis I should select for my proposed study. If I give you some basic information, I'm hoping you can give me some advice on what you would do if it were you so that this study can be completed as efficiently as possibly. I've been at the development/proposal stage of this study now for 2 years, and I'm not sure how much longer I can go around in circles. Any feedback would be extremely helpful!

    Research Question:
    Which prenatal/postnatal environmental risk factor is more prevalent in children with autism spectrum disorder (ASD) than control children? (I'm considering changing the maternal perception of stress and blood lead levels to dichotomous variables, before they were continuous)

    Recruit (a goal of) 75 children ranging in ages from 2-5:11 that have a medical diagnosis or school classification of ASD, match them with 75-typically developing children matched via age, gender, and parental education. Those with ASD can be sampled from a variety of sources in Yuma County- Division of Developmental Disabilities (state-based services), schools, private clinics, support groups. Typicals can be recruited from regular daycare/preschool programs, pediatrician offices.

    IV1- Maternal perception of stress during pregnancy via maternal retrospective recall of stress
    IV2- Blood lead levels obtained via medical release from pediatrician's office
    DV- Absence/Diagnosis of ASD (I'm considering making all three dichotomous variables)


    2. Description of Data Analysis:

    1:1 matched case-control study of offspring (2-5) with ASD. 56 matched pairs. Controls will be age, gender, and parental education (SES)-matched without ASD.



    The goal of the analysis is to determine whether the diagnosis of ASD is related to perinatal/postnatal environmental risk factors (blood lead levels, maternal perception of stress).



    If I use conditional likelihood logistic regression, I can use the information contained in the matches. I wonít be able to estimate the intercept, and with no intercept in the conditional logistic regression model, I canít estimate P. Odds ratio can be estimated using the beta so we can estimate other effects that Iím interested in (besides intercept).



    Conditional Likelihood (Lc)

    Lc = πP(Xi) π[1-P(Xi)] / Σ{ΣP(Xi) π[1-P(Xi)] }

    = πexp(ΣβiXi) / Σ[πexp(ΣβiXi)]



    πP(Xi) =probability of pairís case having the risk factor

    π[1-P(Xi)] = probability of pairís control not having the risk factor

    Σ{ΣP(Xi) π[1-P(Xi)] } = the joint probability that either the case or control

    has had the risk factor.

  2. #2
    Points: 7,920, Level: 59
    Level completed: 85%, Points required for next Level: 30

    Posts
    160
    Thanks
    1
    Thanked 7 Times in 7 Posts

    Re: PhD Dissertation Help- General guidance and direction desperately needed!!!!

    You should try to formulate the problem in statistical language. E.g. the statistical scale of the to predict target variable, the statistical scale of predictors, where to predict from.
    Prediction is very difficult, especially about the future. (Niels Bohr)

  3. #3
    Points: 3,015, Level: 33
    Level completed: 77%, Points required for next Level: 35

    Posts
    47
    Thanks
    8
    Thanked 7 Times in 7 Posts

    Re: PhD Dissertation Help- General guidance and direction desperately needed!!!!


    I would do univariate analysis first on the candidate predictors. You can initial do these by hand or in a stats package. Usually when looking for candidate predictors are more liberal p value is used (such as p 0.1 or even 0.2, rather than usual 0.05). Then once you have your predictors, I would use multivariate regression and develop a model within SPSS. Not sure if that helps?! Good luck

+ 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