one down, how many to go?

[baratron]

Major thesis crisis number 2 has been averted. I was panicking so much, I didn't even get round to telling you about it. To cut a long story short, the reason different people were giving the results of the same method as H-1, 2H-2 and 2H-1 was that they were using different methods! The reason I hadn't noticed was that they are only subtly different - too subtle for a non-statistician to notice at 5am after not enough sleep.

Major thesis crisis number 1 continues to boil. To answer everyone who wondered why I didn't try a book, my problem concerns holes in data. The vast majority of stats books don't seem to consider that data could ever have holes in it. Ordinary stats books only use model data, and applied stats books... well, I went through the entire shelf of books on statistics as applied to the space / atmospheric / meterological sciences in the British Library, and only one of them gave any indication of what you're supposed to do with gaps in your time series. And their method cannot be applied to my data for reasons of chemistry. I looked at a couple of books about statistics as applied to economics (apart from astronomy and meterology, the other biggest user of time series is in economics - such as forecasting the Stock Exchange indexes), but economics is... not my science, and I couldn't make much headway.

As for asking someone at the university, I asked around the maths department 2 years ago when I was actually there most days, and was generally unsuccessful in finding anyone who would admit to knowing anything about statistics. There is a maths consulting service, but it costs money, operates for 2 hours one day a week, and you have to be registered staff or student - I'm only "writing up", which means I have to give the college money and in return get none of their facilities.

Will continue to plug away with the people on sci.stat.edu, and try nitoda's Toni.

Comments

[vampwillow]

It was an *awfully* long time ago but my compsci course did a ton of stats and I did have a couple of standard textbooks on the subject. If I can locate them I'll see what they say about gaps/holes. It was reckoned that we did more maths in our first year and RCS's maths students did!

I do recall a 'pleasant' hour and 10 minutes proving that the number space between '1' and '2' was continuous and thus '1+1=2' was a valid calculation.

eep ... why am I on-line at this time of the night? !!

[rivka.livejournal.com]

The vast majority of stats books don't seem to consider that data could ever have holes in it.

Isn't that bizarre? I went through the same kind of problem with data that didn't conform to the assumptions for the more common analyses. God help you if your data can't be forced to a normal distribution, because the stats gurus sure don't want to.

Does your university have a stats department? Probably more use than the mathematics department.

[lovingboth]

Ah :( I'd try a quick search of other London Uni's prospectuses to find places that do admit to knowing something about statistics, and who. Then ring them.

Is there going to be commercial interest in the results - is there a firm that is going to go 'oooh'? If so, try talking to their R&D people. They may know someone.

[chrismoose.livejournal.com]

Have you tried asking Lynda White in the maths department? She a statician, as well as senior tutor. She was my tutor and is really approachable...

Try emailing her (I'm sure she'd be really understanding if you say you can't actually get into college to ask around, as she has mobility problems herself...).

I'd offer to help, but Stats is really my weakest part of maths... :-(