Hi Joan, I don't know myself, but easy if you just find a manuscript that does what you are trying to do, such as that below. I hope this helps. It is scary the increasing BMI, I see this in teenagers, I think the problem is way too much technology in society, or more so over use and abuse of technology causing sedentariness.  Thanks Debbie

Body mass index has a curvilinear relationship with the percentage of body fat among children.

Federico B1, D'Aliesio F, Pane F, Capelli G, Rodio A.

Author information

1Department of Health and Sport Sciences, University of Cassino, Cassino, Italy. [email protected].

Abstract

BACKGROUND:

Body Mass Index (BMI), which is defined as the ratio between weight (in kg) and height (in m2), is often used in clinical practice as well as in large scale epidemiological studies to classify subjects as underweight, normal weight, overweight or obese. Although BMI does not directly measure the percentage of Body Fat (BF%), it is widely applied because it is strongly related with BF%, it is easy to measure and it is an important predictor of mortality. Among children, age and sex-specific reference values of BMI, known as percentiles, are used. However, it is not clear how strong the relationship between BMI and BF% is among children and whether the association is linear. We performed a cross-sectional study aiming at evaluating the strength and shape of the relationship between BMI and BF% among school-aged children aged 6-12 years.

FINDINGS:

The study was conducted on a sample of 361 football-playing male children aged 6 to 12 years in Rome, Italy. Age, weight, height and skinfold thickness were collected. BF% was estimated using 4 skinfold equations whereas BMI was converted into BMI-for-age z-score. The relationship between these variables was examined using linear regression analyses. Mean BMI was 18.2 (± 2.8), whereas BF% was influenced by the skinfold equation used, with mean values ranging from 15.6% to 23.0%. A curvilinear relationship between BMI-for-age zscore and BF % was found, with the regression line being convex. The association between BMI-for-age zscore and BF% was stronger among overweight/obese children than among normal/underweight children. This curvilinear pattern was evident in all 4 skinfold equations used.

CONCLUSIONS:

The association between BMI-for-age zscore and BF% is not linear among male children aged 6-12 years and it is stronger among overweight and obese subjects than among normal and underweight subjects. In this age group, BMI is a valid index of adiposity only among overweight and obese subjects.

PMID:

21851612

[PubMed]

PMCID:

PMC3169475

Free PMC Article

Deborah, I am asking for a specific syntax in SPSS. I have read the literature!!!! Converting BMI to percentiles for age for a large data set cannot be done by CDC excel formula.

Googling on took me to the links pasted below.  Perhaps you'll find what you need at the second one.

http://www.biomedcentral.com/1471-2431/12/151#B8

http://www.who.int/childgrowth/software/macros_special_spss/en/

PLEASE I need a formula or syntax for SPSS!!!!!

Thank you, Bruce. The program is syntax for SPSS. It is specific to another data set - not NHANES.

Joan, simply click on Analyze, Descriptives, Choose your BMI variable and bring it across to the variables box. Then check the box next to 'Save standardised values as variables. Then click ok. SPSS will automatically generate the Z_BMI variable for you. NO syntax needed. See this youtube video: https://www.youtube.com/watch?v=okf2nokI_tE

The method Sinéad is suggesting will compute a Z-score version of BMI using the sample mean and sample SD of the BMI variable in the active data set.  I may be wrong, but I suspect that is not what Joan is looking for.  (I suspect it involves using age-dependent population means & SDs.)

Joan wrote:  "The program is syntax for SPSS. It is specific to another data set - not NHANES."

Fair enough.  But does that mean it's just a simple matter of changing some variable names in that syntax to match the corresponding variable names in NHANES?  (I don't know that to be the case--I'm just asking!)

Yes, Bruce, it does. I am a 'menu person' and do not like writing snytax. I think it would be easier to use the new NHANES data that already has zBMI as a variable. It is easier to combine cycles and re-run analyses, then to mess with 4 pages of syntax that generate sub-variables.....

I think Sinead's answer will convert Joan's dataset of BMI scores to Z-scores based on her own data, not Z-scores based on a standard reference population.  Sinead's answer will do that in SPSS (it will create Z-scores for her BMI scores based on her study sample only).  But if Joan wants to see what Z-scores her subjects would have relative to an external standard population as if her subjects were members of that population, the CDC has Excel spreadsheets for that.  It's a large task though, involving many spreadsheets for different height and weight limits, ages and genders.    I don't know of any SPSS syntax code that would do this.   

Joan, using the "menus" only in SPSS severely limits you.  As Raynald Levesque put it on his SPSS Tools website, "The need to use syntax cannot be overstated."  See the link below for the many reasons he lists for learning syntax (and note especially the Key Items section).  Finally, note that most SPSS users who do use syntax started with the baby steps of exiting the GUI menus via the PASTE button rather than by clicking on OKAY to run a command directly.  By doing this, you can learn to use syntax at a pace that is comfortable for you.  HTH.

http://www.spsstools.net/LearningSyntax.htm

That link to Raynald's SPSS site is great, Bruce.  Thanks!  I've bookmarked it. 

I understand what you say about learning syntax, but with that said, using the menus is 20 times more convenient --- but obviously they don't do everything you want SPSS to do!  I just wish SPSS (or its supposedly cheaper mimic, PSPP) would put everything on the menus, if this is even possible.  

Thank you Bruce and Jerry! Your answers were very helpful. I also found several universities have websites that are very informative- for example -http://www.ats.ucla.edu/stat/spss/seminars/spss_syntax/default.htm

Did you ever figure out an easy way to calculate BMI-z across a dataset of individuals through SPSS or another method?

There are zBMI tables by CDC. The problem is entering the individuals. This is OK for a few 100 but not for a national data set. Now NHANES has zBMI for children starting in 2013 or 2014. Thanks for asking.

I see, thank you so much for your response. I haven't been able to find complete CDC tables, per say, but rather tables with BMI-for-age for selected z-scores (-2, -1.5, -1, etc). Is there another table to which you are referring that might be useful for a dataset of about 500 kids?

Erin, the only tables I know are the ones provided by CDC and they are age/gender zBMI. They also generate the percentile. It is a problem having a large sample size; however, if you know a statistician who is familiar with SPSS, there is a routine that can do the calculation directly in SPSS using syntax.

hi, if you just need here is your answer

http://www.tulane.edu/~panda3/Analysis2/submods/epicalc/uec_index.htm

Super easy using the menus. 

Analyze/Descriptive Statistics/Descriptives

Put the BMI scores into the variable box and click the check box under the list of variables telling it to save standardised values as variables. Run. Bingo.

Actually Angela, that will produce z-scores based only on her own dataset.  What she's probably looking for is to assign z-scores based on a standard, normative population, and for this she'll have to use the CDC's methodology based on gender, weight, and height. Then she can see where the subjects in her study sample fit in relation to the standard population.  

Aha. Thanks Jerry. That'll teach me not to read carefully :)

one way to do it is using LMS based formula which can be applied after segregating age, sex, and the parameter like height, weight and BMI.

Dear Vaccaro J,

I have recently searched a lot about it, but could not find a suitable inbuilt formula for SPSS. But if you are comfortable with MS Excel, there is a solution. CDC provides their LMS data on their website: https://www.cdc.gov/growthcharts/data/zscore/bmiagerev.xls

You can make an Excel calculator with this LMS data. If you are not interested in making the calculator yourself, then for you, CDC provides the Excel calculator at: https://www.cdc.gov/healthyweight/xls/bmi_group_calculator_metric.xls

You can get an excellent online version of calculator with graph plotting at: https://www.bcm.edu/cnrc-apps/bodycomp/bmiz2.html

Thanks and Regards.

It is recommended that BMIz scores for children up to 2 years should be calculated using the WHO macro (adjusted for gender, age), and the SPSS syntax is available here: http://www.who.int/childgrowth/software/en/

For children over 2 years of age, the CDC macro is preferred, but currently it is only available, on the website, in SAS syntax https://www.cdc.gov/nccdphp/dnpao/growthcharts/index.htm

If anyone has the SPSS syntax let me know (but SAS should be widely available as well)

found this helpful:

https://www.dropbox.com/s/y88smc5ojmz9nv8/bmipct_cdc.zip

link taken from here

http://www.talkstats.com/showthread.php/53437-New-to-SPSS-want-to-find-out-if-there-is-a-way-to-create-BMI-category-variable

Hi could this useful to calculate DHS (demographic n health survey data)?

this is the way to calculate BMI using SPSS:

compute bmi = weight / ((height/100)**2) .

It appears that you are referring to a z-score for BMI. In that case you would need to compare your BMI scores to a standard reference population. The resulting BMI z-score would show you where your subjects would lie in a standard reference population, if they were members of that population. This makes studies comparable to one another. For ages 4-18, the usual reference is the CDC population norm. You can find on-line calculators for converting your scores to z-scores that are standardized to this reference population.