i'd like to know if some of U have even touch this, ie, style analysis...i'used it, i found this pretty funny, and not too stupid...but what about doing a categorisation with it, ie, say that a fund is like a bond, from the united states, with other attributs...or, another fund is a pure fund, belonging to the sectors of the pharmaceuticals...???...any opinion will be great...

QuoteOriginally posted by: figarchi'd like to know if some of U have even touch this, ie, style analysis...i'used it, i found this pretty funny, and not too stupid...but what about doing a categorisation with it, ie, say that a fund is like a bond, from the united states, with other attributs...or, another fund is a pure fund, belonging to the sectors of the pharmaceuticals...???...any opinion will be great...It is not too stupid if you want to show that a 100% S&P500 index fund actually is not, but it is not a DNA analysis and you will never find who are the fathers of a given return exactly.

well...for a index fund, that's sure...but, if we'd like to buy a pure fund, that is, which is invested in only one sector, like "or", "biotechnology", is it not interesting to know that it is a really pure fund...???...another thinking, if we categorize all the funds with this mean, can we imagine that the results will give another categorization, different from morningstar or s&p, which will be more interesting for fund of fund...in fact, we can may be have funds'stock which are far from their benchmark, but belong on the top for the growth value...or their exposure is the emerging market index...

I think the ambiguity that Mukuzani refers to stems from the fact that style models are usually factor models, with "reference" or benchmark portfolios built out of "pure" funds of a given style. For a given test fund, you can "decompose" the returns into components of each style. Different style analyses will come up with different decompositions, because it's so hard to build pure style portfolios that aren't highly collinear, and the factor design is itself somewhat arbitrary. I've seen it used as an informal management tool, rather than anything rigorous.

its pretty useful for cash equity trading in markets where short-selling is NOT allowed. the important thing is even if two baskets have a correlation more than 0.30 then they will always fall or rise together over a period of 1 week in a bear or bull market. i have usually found correlations less than 0.30 to be a good cut-off. in bear markets, its pretty hard to make money on broad market indices or large market cap stocks, except in a bear market rallies. one way out is to try and create indicies usuing criterias like capitalization, growth/value and sectors and take a view on them. sector funds are usually also not that effective, unless you go for low market-cap sectors, for instance sectors like IT, Pharma, etc are likely to have a correlation more than 0.30 with the broad market index and are no good. on the other hand, sectors like heavy engineering, cement, or fmcg etc can be good. the key is to avoid sectors with too many stocks in the broad market-cap indices, say for instance stocks in the S&P 100. another factor could be use dividend yield, but didn't find it useful. choose sets of indices that have less than 0.30 correlations (say over a period of 1 previous bull and 1 previous bear market).

QuoteOriginally posted by: pb273its pretty useful for cash equity trading in markets where short-selling is NOT allowed. ).It is true if the goal is style analysis itself. If you are going to use weights for portfolio optimization, keep in mind that the residuals of a constrained regression are not independent in general. As I remember in the original paper by W. Sharpe this problem is not mentioned.

strange is your remark...in a first time, i was agree with you, but now,...i've done a run test, a von newman test and a turning point test, same results...they seemed to be independant...may be because my data are monthly, or, i don't, know...strange is it...because i've a lot of variables...???...