Quote:
Simply being called 'fat' makes young girls more likely to become obese: Trying to be thin is like trying to be tall.
Okay, I'll buy that, now why are they saying it. So we read along, link to the summary page, the download the eventual article. Firstly it's a page and a half long letter, with a single table, which is worrisome in itself. Below is the table in which they display their results.
Source of Labeling, OR (95% CI) Predictor Model 1: Anyone Model 2: Family Model 3: Nonfamily Baseline BMI 1.70 (1.61-1.80) 1.70 (1.61-1.80) 1.72 (1.62-1.82) Race 1.31 (0.93-1.84) 1.30 (0.93-1.82) 1.32 (0.94-1.86) Parental education 0.73 (0.58-0.93) 0.73 (0.58-0.93) 0.75 (0.59-0.95) Household income 0.76 (0.64-0.89) 0.76 (0.64-0.89) 0.74 (0.63-0.88) Age at menarche 1.01 (0.91-1.12) 1.00 (0.90-1.11) 1.01 (0.91-1.13) Baseline labeling 1.66 (1.20-2.30) 1.62 (1.18-2.22) 1.40 (1.01-1.94)
Important notes:
1) Higher numbers mean a better correlation
2) The 95% confidence interval is range of values possible given the sampling size, scatter, etc.
The couple of things that irked me are that firstly there is probably a better correlation between baseline BMI (age 10), and final BMI (age 19) than being labelled as fat prior to the study. So first off it's completely within the data to conclude, the better correlation between being fat at 19 is being fat at 10, rather than someone calling you fat. You could even conclude that because the correlation value is lower calling someone "fat" might help lessen their chances of being overweight. Of course all the uncertainty leads me to point number 2 below.
Secondly, the error range is so broad you can't really conclude anything from the data. Seriously, if your confidence intervals overlap, you're supposed to assume you can't tell the difference. Baseline labeling and Baseline BMI should be considered indistinguishable as to which is a better predictor of BMI at the end of the study. This is like community college statistics here people!
Finally there's no division of the data on Baseline BMI, so there's way to know if people are calling a person "fat" because they're overweight initially, or because they're being evil jerks Not that those are mutually exclusive mind you. Every experiment needs a proper control, how this one didn't have one worked into the data is beyond me. You need to control for known causes in your analysis, and childhood BMI is known to be well correlated with adult BMI.
I'm mean it's like some grad student threw a bunch of data at a SPSS and it didn't pan out. So they had to dream up something to publish because otherwise there was nothing worthwhile to show for a 9-year study. It's not that this isn't real, or isn't a problem, but doing a poor study doesn't help advance anyone's understanding of the issue. All you're really showing with the data is that people call heavier people "fat," and we already knew that.
Seriously, do a proper experiment or GTFO.
So what do you think?
| I think someone pissed in your cheerios or something.: |    | 3 (27.3%) |
| It's obviously a poorly designed study.: | | 4 (36.4%) |
| No, you're reading the data wrong stupid. Here let me show you how this works...: | | 3 (27.3%) |
| I have no idea WTF you're talking out.: | | 1 (9.1%) |
| Total: | 11 |
Edit: Okay I feel better now...
Edited, Apr 29th 2014 2:54pm by someproteinguy
