Enquiry about statistical test

Rich Ulrich

2018-05-16 06:06:47 UTC

Post by l***@gmail.com
Hi. I do want to ask, is there possible to run any statistical test based on the table below?
I would like to see the association between the healthy score of the place and the BMI.
Thank you very much.
Variables Healthy Scores BMI (kg/m2)
Boys Girls Boys Girls
Mean (SD)
place A 14.67 (22.62) 14.26 (21.83) 18.18 (5.31) 16.83 (3.14)
place B 24.91 (72.00) 23.88 (68.98) 16.05 (3.14) 17.04 (5.56)
place C 13.65 (30.13) 13.11 (28.93) 19.48 (4.06) 19.01 (2.59)
place D 3.22 (8.45) 3.10 (8.11) 18.84 (4.11) 21.34 (4.16)
place E 6.52 (17.32) 6.26 (16.63) 17.85 (3.54) 20.75 (5.33)
Total 11.08 (31.85) 10.64 (30.55) 17.93 (4.31) 18.84 (4.76)

1. The ordinary ANOVA table requires the Ns for each cell.

The usual analyses cannot be performed on means and SDs alone.
SPSS can input a matrix of means, SDs, and Ns for procedure
Oneway; I don't recall how you can use those elsewhere.

But I have serious questions about the data, so I'm not going
to look that up.

2. The so-called Healthy scores are weird.
Boys 15 25 14 3 7
Girls 14 24 13 3 6

What is that set of scores?

The scores are greatly skewed (it seems), with SDs constricted
when near to zero; and the SDs all are greater than the means.
That tells me that the raw scores probably should be transformed
in some fashion, if they are going to serve very precisely.

Despite that initial measurement problem,
- the places vary a lot (by effect size) between places.
- places show a correlation for the 5x2 points of r= .99+.

It looks like Healthy Score is an artifact of (or otherwise closely
tied to) Place, in order for the correlation between sexes to be
so strong.

3. BMI means do fall in a range for children. The one time that I
worked with BMI for children, I was provided standardization
by ages, if I recall correctly. Maybe that accounts for some oddity
in BMI reports.

For two particular groups -
BMI Boys Girls
Place A 18.2(5.31) 16.83(3.14)
Place D 18.4(4.41) 21.34(4.16)

- I see Boys are practically identiical; Girls differ by more than
one SD - a large effect size - skinnier than boys in A and fatter
in D.

- Another question about simple correlations with BMI is
the assumption that this should be considered a monotonic
relationship, i.e., one end is good and the other is bad ...
or else, one extreme end does not exist in the samples.

Unless you are worried solely about starvation or about
obesity, you should not be comparing means as your
most interesting test.

--
Rich Ulrich