The handedness survey http://www.parkinsonsmeasurement.org...ess/survey.htm
has thrown up an interesting result unrelated to handedness.

Data from the survey suggests the time beween first presenting with Parkinsonian symptoms to a doctor and the diagnosis of IPD being given was shorter for men than for women. On the sample studied, the average for men was 0.55 years, whereas for women it averaged 2.62 years.

It is interesting to speculate as to why this may be the case. The sample, the result of a survey posted on this and other fora, is undoubtedly biased to early onset PwP. No claim is made that the average age of diagnosis for men (48.45 years) and for women (51.25 years) are representative of the whole IPD population. But they are sufficiently close to think that there is no internal bias on the question of who gets diagnosed fastest.

It is thought that in general women seek medical advice more promptly than men [1]. In the case of IPD this could mean seeking medical advice before the symptoms were developed enough for a doctor to diagnose IPD. Assuming the aetiology is the same for men and women, this suggests that for men there is on average a post-symptomic, albeit insufficiently pronounced for them to seek medical advice, period of at least 2 years before medical intervention is sought.

What follows is a justification of the statistics. Hopefully, the account is detailed enough for someone familiar with the area to spot weaknesses in my
analysis.

The results are based on those records available when the analysis started (the first 87 records).
For the raw data see:
http://www.parkinsonsmeasurement.org...ss/results.xml
For a statistical summary see:
http://www.parkinsonsmeasurement.org...ss/summary.htm
Three records were not for IPD, and were discarded.

A data mining, post hoc approach was adopted: the data was scanned for apparent anomalies (rather than being used to confirm or reject a prior conjecture). Left unadjusted this would give misleadingly large numbers of significant results. Talking roughly, if you randomly test 100 things, chance will bring up 1 event that has probability less than 0.01. To reduce this effect the data was split into two groups: the first group of 30 records was used for data mining, and then discarded; the second group of 54 records (22 men, 32 women) was used to test for significance.

The second group were distributed as follows (A=years between first presentation and diagnosis, B = # men, C = # women)
A 0 1 2 3 4 5 6 7 8 9 10 >10
B 15 4 1 2
C 15 2 4 1 3 1 0 2 1 1 1 1@13

A t-test comparing the means for men and women gives t=3.15, with df=62. The single tailed 0.01 critical value is 1.578, giving a 99% confidence interval for the population difference of [0.501, 3.658].

It is not clear that either group is Normally distributed. Moreover, the variance of the two gender groups differs. So a non-parametric approach based on ranking was also used. Put simply, if the two groups were from the same population, one would expect the average ranking (in a combined list of the two groups) to be the same - 27.5/54. In fact, on average men were ranked 22/54 and women 31/54. The Mann-Whitney test was used to generate a U value of 238.5, compared to an expected value of 352. The U value is approximately Normally distributed [2]. A single tailed significance test gives a z value of 1.998 and p<0.03 for the null hypothesis that the time is at least as long for men.

[1] http://www.nlm.nih.gov/medlineplus/menshealth.html
[2] http://en.wikipedia.org/wiki/Mann%E2%80%93Whitney_U

John