#### Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Options

# Blog - El Niño project (part 7)

I've started writing a new post in this series:

It's a bit dry so far, mainly about the definition of El Niño. But I may also include a bit about "different flavors of El Niño". I will send a bunch of an article about that - Graham Jones already has it. If you don't get it from me in a few minutes, and want it, let me know.

• Options
1.

Okay, it's ready now - check it out! It's pretty unambitious, but I think that's okay: it does a limited job that needs to be done, and we can move on to more interesting things next time.

Comment Source:Okay, it's ready now - check it out! It's pretty unambitious, but I think that's okay: it does a limited job that needs to be done, and we can move on to more interesting things next time.
• Options
2.

I've had a look and it looks good (as you say given the subject). My only comment is that the ending is very abrupt, perhaps a final sentence?

Comment Source:I've had a look and it looks good (as you say given the subject). My only comment is that the ending is very abrupt, perhaps a final sentence?
• Options
3.

Okay, I was getting tired, but I'll add a better ending and then post this.

Comment Source:Okay, I was getting tired, but I'll add a better ending and then post this.
• Options
4.

I added some fun stuff at the end, mentioning neural networks as a way to classify different "flavors" of El Niño.

Comment Source:I added some fun stuff at the end, mentioning neural networks as a way to classify different "flavors" of El Ni&ntilde;o.
• Options
5.

I hope I didn't goad you into doing all that extra writing. Regardless the finished blog article is very impressive.

Comment Source:I hope I didn't goad you into doing all that extra writing. Regardless the finished blog article is very impressive.
• Options
6.

Thanks! I don't mind goading; the ending was indeed quite dull before, the product of getting tired rather than any plan. It's better to dangle some "coming attractions" in front of the reader.

Comment Source:Thanks! I don't mind goading; the ending was indeed quite dull before, the product of getting tired rather than any plan. It's better to dangle some "coming attractions" in front of the reader.
• Options
7.

It looks fine to me, I actually like the RED BLUE number highlights.

I prefer these short blog articles to learn from and get ideas for coding

Dara

Comment Source:It looks fine to me, I actually like the RED BLUE number highlights. I prefer these short blog articles to learn from and get ideas for coding Dara
• Options
8.
edited August 2014

Hello John

I hope you propagate the concept of backtesting for 100s of months of data for these sorts of data and forecasts. Also the accuracy measurements for output that requires IF THEN ELSE e.g. sign of some number or some number being above some value, requires double stats for IF and ELSE separately. Since most forecast algorithms are lop-sided towards IF or ELSE. For example to forecast that an index number is more than 0.5 has different accuracy than forecast accuracy for less than 0.5.

To claim success that a forecast algorithm predicted something last 2 times or 3 times, is laughable in most markets e.g. if I told wall street traders that my forecast algorithm predicted the price of IBM stock past 2 days with good accuracy they will laugh their heads off.

I have self-imposed backtesting on my computations but I suggest that we do it for all

Dara

Comment Source:Hello John I hope you propagate the concept of backtesting for 100s of months of data for these sorts of data and forecasts. Also the accuracy measurements for output that requires IF THEN ELSE e.g. sign of some number or some number being above some value, requires double stats for IF and ELSE separately. Since most forecast algorithms are lop-sided towards IF or ELSE. For example to forecast that an index number is more than 0.5 has different accuracy than forecast accuracy for less than 0.5. To claim success that a forecast algorithm predicted something last 2 times or 3 times, is laughable in most markets e.g. if I told wall street traders that my forecast algorithm predicted the price of IBM stock past 2 days with good accuracy they will laugh their heads off. I have self-imposed backtesting on my computations but I suggest that we do it for all Dara