that you are over-fitting the model, because there are so many free variables (and by chance some could correlate). I definitely need to take a closer look at the stats, but the work looks good so far. Of course it must be because it fits my world view! ;-) Per Alm would be happy to see that these results seem to fit his experimental findings on a smaller sample. Lets look at the various ones:
Premature birth: involves some damage to brain,
attention deficit: too or some neurotransmitter level issues,
alcohol abuse: not sure,
OCD: not surprising,
disabled mother: strange,
foreign parent: would support findings that bi-lingualism could be a risk factor.
These findings are of course not very helpful in treatment: many causes (or just correlations) possible, but it might be that these are symptoms of some brain damage in general. And the vast majority of kids with premature birth do not start stuttering but this subgroup is more likely to start stuttering. Please also note that some factors are clearly not genetic like foreign parent but other could be more genetical like OCD.
Here is the 2009 October 15 article from Ajdacic-Gross, Vetter, Müller, Kawohl, Frey, Lupi, Blechschmidt, Born, Latal, Rössler in Eur Arch Psychiatry Clin Neurosci:
The spectrum of risk and concomitant factors in stuttering is generally thought to be wide and heterogeneous. However, only a few studies have examined these factors using information from large databases. We examined the data on 11,905 Swiss conscripts from 2003. All cases with high psychiatric screening scores indicating "caseness" for a psychiatric disorder were excluded, among them potential malingerers, so that 9,814 records remained. The analyses rely on self-reported information about stuttering in childhood, problems at birth, problems in school, mental disorders of parents and relatives, childhood adversity and socio-demographic information. Statistical modelling was done using logistic regression and path analysis models. Risk factors determined in the logistic regression include premature birth, probable attention deficit hyperactive disorder, alcohol abuse of the parents, obsessive-compulsive disorder in parents and relatives, having a disabled mother and having a parent from a foreign country. There is no overwhelmingly strong risk factor; all odds ratios are about 2 or below. In conclusion, large databases are helpful in revealing less obvious and less frequent risk factors for heterogeneous disorders such as stuttering. Obviously, not only secondary analyses, but also systematical large scale studies would be required to complete the complex epidemiological puzzle in stuttering. An extensive examination of young adults who were initially assessed in childhood might provide the most promising design.
No comments:
Post a Comment