« of social moods and investments in advertising, 2 | Main | Decisions, data, and interpretations »

05 November 2009

Comments

Neil Charles

Nice post. I always think it's fun explaining to clients that if they want to explain the past (and work out ROI) they're going to need one model. For accurate forecasting they'll need a different one...

Response is almost always 'but surely you only need one model if it's right?'

Philip Tiongson

Thanks very much for the comment. Indeed, fun! Although sometimes, it could painfully be one great fun!

What's worse really is when people rely too much on "tools that automate the model selection process". To which I typically say, yeah: saves me perhaps some time by consolidating computer power - BUT it does not replace the thinking process which ultimately I will have to do any way - software or no software.

Interpretation is still best left to us - and yes, just because we got the past explained doesn't mean we've got the future locked on. :D

The comments to this entry are closed.

Twitterlogs

    follow me on Twitter
    Related Posts Plugin for WordPress, Blogger...

    Empire Avenue

    • I am Listed on Empire Avenue

    Sites that interest me

    Blog powered by Typepad
    Member since 10/2006