+ Reply to Thread
Results 1 to 5 of 5

Thread: Longitudinal analysis

  1. #1
    Points: 16, Level: 1
    Level completed: 31%, Points required for next Level: 34

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Longitudinal analysis




    Hello. I would love some help:

    I have a dataset with 111 cases of a rare disease registered throughout a 10-year period in one Hospital. I have something like 50 variables (most of them are dichotomous, but some are ordinal and some are continuous intervals).

    Initially, I divided this 111 cases in 2 groups (those diagnosed from 5 to 10 years ago, and those diagnosed in the last 5 years) and I found several peculiarities and epidemiological changes between these 2 groups.

    My question is: is there a more elegant analysis that I could do, like analyzing the data longitudinally, instead of dividing it in 2 groups?

    I can handle SPSS (and syntax) if you give me some general orientations. However, I currently donít possess sufficient knowledge and time to learn R statistics (or other software of this kind).

    Thank you!
    Lucas

  2. #2
    Points: 2,462, Level: 30
    Level completed: 8%, Points required for next Level: 138

    Posts
    200
    Thanks
    20
    Thanked 48 Times in 43 Posts

    Re: Longitudinal analysis

    Hi, I don't know your research questions in detail, but how about introducing "year" as a (probably centered) continuous variable, so that you can investigate the change of these variables in time? Or how about analyzing interaction terms between "year" and other variables?

  3. #3
    Points: 4,664, Level: 43
    Level completed: 57%, Points required for next Level: 86
    kiton's Avatar
    Location
    Corn field
    Posts
    234
    Thanks
    47
    Thanked 51 Times in 46 Posts

    Re: Longitudinal analysis

    Depending on the nature of your dependent variable, you can specify a panel (i.e., longitudinal) model. For that, you will need each of your 111 cases to have a unique ID and also have a date variable pertaining to your observations. Once you declare your data set as "long", you'll be able to estimate fixed- or random-effects models. Sorry, I cannot help you with SPSS but I am sure Google knows a ton of sources.

  4. #4
    Omega Contributor
    Points: 38,374, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    6,998
    Thanks
    398
    Thanked 1,186 Times in 1,147 Posts

    Re: Longitudinal analysis

    So we don't mislead you, please better describe the setting, research questions, disease progression, etc. I wonder about discontinuity regression , know little about the data and research design.
    Stop cowardice, ban guns!

  5. #5
    Points: 16, Level: 1
    Level completed: 31%, Points required for next Level: 34

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Longitudinal analysis


    Hi! Thank You all for answering!

    Quote Originally Posted by mmercker View Post
    Hi, I don't know your research questions in detail, but how about introducing "year" as a (probably centered) continuous variable, so that you can investigate the change of these variables in time? Or how about analyzing interaction terms between "year" and other variables?
    Well, I don't know if I can do this, but one of the main findings of my data is that the time (in days) required for diagnosis of this disease has shortened considerably throughout time (and the severity of the disease at presentation also). Apart from medical experience (witch I believe accounts considerably for these outcomes), I believe that some other factors (like associated comorbidities, age and risk factors) may be playing a major role in this change (because they are also changing over time).

    Yesterday I recoded the variable "date of diagnosis" to a continuous variable (like: 2006=1; 2007=2; 2008=3 ... 2016=10). Then I ran some correlation tests (spearman) and multiple regression. My dependent variable was time required for diagnosis (in days). The independent variables were: the variable "years" (that I recoded), age, and several dichotomous variables that made sense put in the equation. In the regression, I found that some of this variables were good predictors of the dependent variable (the most important being the variable "years" that I recoded).

    Can I trust this results?

    Quote Originally Posted by kiton View Post
    Depending on the nature of your dependent variable, you can specify a panel (i.e., longitudinal) model. For that, you will need each of your 111 cases to have a unique ID and also have a date variable pertaining to your observations. Once you declare your data set as "long", you'll be able to estimate fixed- or random-effects models. Sorry, I cannot help you with SPSS but I am sure Google knows a ton of sources.
    I will search for sure! Thanks!

    Quote Originally Posted by hlsmith View Post
    So we don't mislead you, please better describe the setting, research questions, disease progression, etc. I wonder about discontinuity regression , know little about the data and research design.
    So, basically we registered every case of a relatively rare disease that was diagnosed in our hospital in the last 10 years. The variables were collected during hospital admission. There are data of all kind: age, risk factors (Yes/No), Symptoms, Labs, Radiological Findings, Ordinal scale regarding severity of disease at presentation etc.

    While a lot of analysis can be done, I`m particularly interested at the moment in knowing if there is a explanation (other than medical expertise) that could explain why patients are being diagnosed early (number of days from the first symptom to diagnosis) and why the clinical presentation is becoming less and less severe (ordinal scale).

    What i`ve done:
    - Multiple regression (dependent variable --> time required for diagnosis) (and one of the independent variables was "Years" that I recoded from the variable date of diagnosis (described it above).
    - Ordinal regression (dependent variable --> disease severity) --> used the procedure PLUM in SPSS

    Thank You all!

    Lucas

+ Reply to Thread

           




Tags for this 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