Halifax -1.9% MoM, -17.5% YoY -21.3% from peak.

Wookster

Sir_Humphrey wrote: »

It is the size of the sample, not the sample proportional to population that matters.

Actually its not really the size of the sample that matters, its what you do with it that counts.

Seriously though its how the sample is selected. I would say that 1,000 chosen at random out of 25,000 may give an approximation to the entire population if the population is distributed normally.

StevieJ

stephen163 wrote: »

Using MoM data in that way is folly anyway! They should be shot for poor journaism

Yes they should use Qtr on Qtr that has shown a significant reduction in falling house prices for Q1

·[FONT=&quot] [/FONT]House prices declined by 2.7% in Quarter 1. This was smaller than the 5-6% falls recorded in each of the three previous quarters.

stephen163

Really2 wrote: »

Ok then I will let you pick 1000 numbers from 25,000,000 to find the 1000 I am thinking of.;)

Lets put it this way to you guys.

Do you think every house went up 1.9% earlier this year?:rolleyes:
Does a year give a better idea than a month?:rolleyes:

Low samples do not give an acurate idea of the market what so ever that why the odd spike or massive drops are happening.

100,000 houses sold gives a lot better indication than 1,000 if you belive otherwise I worry for the country.

Read a statistics book! Nobody has said that 1,000 is a better indication than 100,000. Also, if you think of 25,000,000 random numbers, sampling 100 numbers would be useless - however, if your numbers were not random, but had a mean and standard deviation, then sampling would work!

Sir_Humphrey

StevieJ wrote: »

Yes they should use Qtr on Qtr that has shown a significant reduction in falling house prices for Q1

·House prices declined by 2.7% in Quarter 1. This was smaller than the 5-6% falls recorded in each of the three previous quarters.

The quarterly figures are still affected by seasonality, so the YoY figures are best, although of course they lag a lot.

Really2

Sir_Humphrey wrote: »

They could have done, but the probability would probably be less than 0.01%. On the other hand they could have gone down by more than 1.9%. I do not know the exact margin of error.

A margin of error by any amount when doing financial acounting or using data for financial reporting can be disasterous (4% for a bank would be the end in the boom if the wrong way).

The current situation kind of proves it don't you think.

Really2

stephen163 wrote: »

Read a statistics book! Nobody has said that 1,000 is a better indication than 100,000. Also, if you think of 25,000,000 random numbers, sampling 100 numbers would be useless - however, if your numbers were not random, but had a mean and standard deviation, then sampling would work!

To what degree of accuracy?

I never said useless but I clearly said the smaller the sample the smaller the accuracy.

Sir_Humphrey

Really2 wrote: »

A margin of error by any amount when doing financial acounting or using data for financial reporting can be disasterous (4% for a bank would be the end in the boom if the wrong way).

The current situation kind of proves it don't you think.

Which is why banks need enough reserves to allow for those errors.

Wookster summed it up nicely. I can only reiterate what Stephen163 said - read some stats books because if you are trading without knowing this stuff you could end up seriously broke.

neas

MItcha ... Updated my sig for you .

geoffky

too many variables around at the moment to have confidence in anyone's figures...too few sales ect ect.... over .3 months should be a better marker...

stephen163

Really2 wrote: »

To work out a companies credit rating do you look at that company actual accounts or 1000 UK companies and average it out.

No, but if you wanted to work out the AVERAGE credit rating for a medium sized company in the UK, you could take a sample and get an estimate.

With house prices, there are many different factors (location, garden, type of house etc.) so they weight each factor to work out the average influence of each factor, then apply this to a house exhibiting average charateristics. You CAN do this and get a good estimate and it is surprising how low the sample size can be to get a good estimate.

You couldn't do this with something random though, like numbers plucked out of thin air or company accounts numbers.