Discussion: View Thread

  • 1.  Creating a size variable based on length/weight

    Posted 01-22-2014 19:54
    Hi All,

    I am trying to build a linear model or logistic model to answer a question like "What are the factors during pregnancy that affect the size of the babies?". Based on the data available, I have two variables "weight" and "length" which are related to size. I need somehow combine these two variables to determine a new variable called size. The one that is most obvious to me is classifying a baby as small if his/her length and weight are both below the average. But this method seems a little bit inappropriate. I am wondering if there is any more appropriate method that I can adopt? 

    Thanks,

    -------------------------------------------
    Yuanyuan Fu
    -------------------------------------------


  • 2.  RE:Creating a size variable based on length/weight

    Posted 01-22-2014 19:58
    There are many ways to do this.

    One would be to turn both height and weight into z-scores (standardize them) and then add them. 

    Another would be to do factor analysis (even on just 2 variables) and use factor scores. The results would probably be very similar to the former method.

    Peter



    -------------------------------------------
    Peter Flom
    -------------------------------------------








  • 3.  RE:Creating a size variable based on length/weight

    Posted 01-22-2014 20:35
    Hi Yuan,

    Along with weight and length, it is also critical to consider gestational age.  There are normative reference values (percentiles) for size for gestational age (see below).  There may be more recent references but this will give you a place to start.  

    Infants 24+ weeks gestational age:  Dombrowski MP, Wolfe HM, Brans YW, Saleh AA, Sokol RJ.  Neonatal morphometry: relation to obstetric, pediatric, and menstrual estimates of gestational age. Am J Dis Child 1992;146:852-6.  

    Infants < 24 weeks gestational age: Klaus MH, Fanaroff AA. Care of the high-risk neonate. Philadelphia (PA): WB Saunders; 2001. p. 632.

    I used those references when analyzing pediatric cohort data (children born between 1988-1993):  Grischkan, Storfer-Isser, Rosen, Larkin, Kirchner, South, Wilson-Costello, Martin, Redline (2004).  Variation in childhood asthma among former preterm infants.  Journal of Pediatrics, 144, pp 321-326.

    -------------------------------------------
    Amy Storfer-Isser, Ph.D.
    Statistical Research Consultants, LLC
    -------------------------------------------




    Original Message


  • 4.  RE:Creating a size variable based on length/weight

    Posted 03-12-2014 13:59
    CDC has a set of percentiles that are in common use as does WHO.
    Different assumptions and different numbers.  The WHO numbers seem to
    be above the CDC numbers.  by about 5 percentile points (plus or minus).
    Ray.

    -------------------------------------------
    Raymond Hoffmann
    Professor
    Medical College of Wisconsin
    -------------------------------------------






    Original Message


  • 5.  RE:Creating a size variable based on length/weight

    Posted 01-22-2014 19:59

    Hi Yuanyuan,

    There are charts used by pediatricians etc. that are like BMI charts, but they indicate percentiles.  Here is one for boys that I found online. I am sure there are more official ones out there, but you can see what I am talking about:

    http://images.meredith.com/parents/pdf/BMI_boys.pdf

    Best,

    Elaine

    -------------------------------------------
    Elaine Eisenbeisz
    Owner and Principal Statistician
    Omega Statistics
    -------------------------------------------








  • 6.  RE:Creating a size variable based on length/weight

    Posted 01-23-2014 08:21

    If you have enough cases, consider using a set of variables in a MANOVA-type  or canonical correlation model.



    -------------------------------------------
    Arthur Kendall
    Social Research Consultants
    -------------------------------------------








  • 7.  RE:Creating a size variable based on length/weight

    Posted 01-23-2014 09:00
    You've probably already thought about this, but as a general approach, I would be leery about integrating these without strong theoretical mechanisms. Length and weight may be deferentially affected by different "predictors".

    That being said, assuming that's been considered, I think you have a couple options:
    1) Easy - BMI is the typical integration of these factors...a quick search turned this up: http://www.ncbi.nlm.nih.gov/pubmed/19160899
    2) Harder - build your own following my initial comment. What is the construct to be represented? What is the biology supposed to be doing? Are we accelerating growth? Are we depleting resources? Are we affecting skeletal growth? Crawl inside your collaborators' heads, extract bits of their expertise that can be woven into theoretical models and proceed.


    -------------------------------------------
    Jason T. Machan
    Director, Lifespan Biostatistics Core,
    Lifespan Hospital System
    Research Scientist, Biostatistics, Research
    Rhode Island Hospital
    Assistant Professor, Departments of Orthopaedics and Surgery
    The Warren Alpert Medical School, Brown University
    Director Biostatistics Externship, Adjunct Assistant Professor, Department of Psychology
    University of Rhode Island
    -------------------------------------------






    Original Message


  • 8.  RE:Creating a size variable based on length/weight

    Posted 01-23-2014 09:17
    You'll find plenty of theoretical support and prior studies in the literature on allometry--anything from evolutionary development to modern entomology and ecology.  Common approaches include principal components and Model II regression (MA or RMA).  As has been mentioned, the best starting places for child growth/health data are CDC and WHO.   http://childhealthdata.org/  , maintained by HHS and NCHS, would be a good start.

    Dave


  • 9.  RE:Creating a size variable based on length/weight

    Posted 01-23-2014 10:39
    What about using percentiles from growth charts?  Here's a link to a growth chart of boys, age 2-20 years.  I'm told there are similar ones for infants.

    http://www.chartsgraphsdiagrams.com/HealthCharts/growth-2-20-boys.html

    -------------------------------------------
    Wayne Fischer
    Statistician
    University of Texas Medical Branch
    -------------------------------------------






    Original Message


  • 10.  RE:Creating a size variable based on length/weight

    Posted 01-25-2014 21:29
    For newborn babies, there is a BMI-like index called the Ponderal Index that people I work with have used to identify babies who were Small for Gestational Age (SGA).  See Kiefer et al, Delayed maturation of auditory-evoked responses in growth-restricted fetuses revealed by magnetoencephalographic recordings. Am J Obstet Gynecol. 2008 Nov;199(5):503.e1-7.  (PubMed PMID: 18533116).  Check out the article, and also check out references 5 through 8 in the article's references. 

    -------------------------------------------
    Eric Siegel
    Biostatistician
    Univ of Arkansas for Medical Sciences of Biostatistics
    -------------------------------------------