House prices fall 1.8% in feb

Swipe

Why is it all these index threads always go off topic and turn into slanging matches?

ad9898_3

Swipe wrote: »

Why is it all these index threads always go off topic and turn into slanging matches?

mmm, I don't really see the chucky/Dan/downey stuff as slanging, I think it can be quite funny in a light hearted banter sort of way.;)

Cat695

mr.broderick wrote: »

Where are you?

only over in France...Peronne

doing research on the battle of the somme (mainly derville wood) and the battle of flers corcelette (the first time tanks were used....remember i'm a calvary soldier)

mr.broderick

Cat695 wrote: »

only over in France...Peronne

doing research on the battle of the somme (mainly derville wood) and the battle of flers corcelette (the first time tanks were used....remember i'm a calvary soldier)

Went to caen a few years back, enjoyed it.

GDB2222

lukekelly wrote: »

As far as I'm aware that is what they're doing. There are problems with their models, particularly their habit of using a 30 year period for the trend-line which means that strong price movement 30 years ago can alter the trend-line substantially as old quarters drop off. They do though use all the standard regression techniques, and certainly don't project backwards when creating these graphs. You can actually download their raw data if you so wish and construct your own trend-lines, you'll find you come up with something very similar.

Negative RPI shouldn't make the graph go haywire. It's cumulative. A year of it will have limited affect on the cumulative inflation figures, at least in terms of long-term trend-lines. The trend-line will certainly be moving downwards in response to the current changes though, two years of 15% falls on a 30 year data series should knock off a good fraction of a percent.

Thanks, I will. BTW, I said polynomial, but I meant exponential. My brain is getting very tired.

One of the problems with time-series analysis is choosing the period to analyse over. I cannot see any really good reason for excluding any part of the last 50 years, as reasonably good data has been collected since the mid-1950s. To the extent the data is cyclical, one really wants to include whole cycles, so an arbitrary 30 year period is not helpful.