From

Location

Message

Date Posted

SF Bay Area, California

This is a follow-up to my earlier post on probability predictions. I've updated my analysis to help correct for the bias in the results caused by the lack of available data on pumpkins under 500 pounds.

Using the 2.5 sigma point as the maximum probable weight predictor, the results are correlating nicely with known outcomes on seeds lines that are nearly planted out.

Here are the top five proven seeds based on the maximum probable weight potential.

Seed Line Max. Probable Wt Top Offspring
1. 1068 Wallace 1,590 1333 Connolly
2. 842 Eaton 1,461 1446 Eaton
3. 1370 Rose 1,456 1273 Howard
4. 1097.5 Beachy 1,452 1438 Carlson-Petersen
5. 723 Bobier 1,441 1407.3 Wolf

A more complete listing of the top ten can be found at:

http://home.pacbell.net/diana_do/probability.htm

Out of the top ten proven seeds, only the 1068 Wallace shows a reasonable probability of popping a 1,500+ pumpkin. The probability of this seed line popping a 1,500+ is 1.8% or about one in fifty.

The 1370 Rose had less than 20 data points so there is less confidence in its results. I have a feeling that its ranking will increase after more data is obtained.

1/10/2006 4:36:45 PM

Ctpumpkin@optonline.net

You've done some thorough & thoughtful work here Vince. Well done.

1/10/2006 6:15:47 PM

Mn

Like wise.

1/10/2006 7:37:52 PM

Mulino, Oregon

How many people in here think Vince needs something else to do this winter? Vince I am totally humbled at the stuff you are coming up with. It has been a long winter, hasn't it? :o)

1/10/2006 8:34:24 PM

Eugene,OR

Guys like Vince make up stuff like this during commercials while watching shows like "The 1000 Greatest Science Discoveries Ever." Then they take an extra five minutes during their lunch break and whip up a little web page, so the rest of us can try to understand it.

1/10/2006 9:58:59 PM

Pumpkintown, SC

LOL, Vince IS the man, BF

1/10/2006 10:07:32 PM

Austria

Wow, that IS interesting. Makes me consider planting the 1097 remake.

1/11/2006 3:12:18 AM

Sioux Falls, SD garyboer@dakotalink.com

Yeah. . .I think this is a better approach
thanks Vince (Diana help with this?)

1/11/2006 9:36:44 AM

SF Bay Area, California

Winter is tough on someone who likes to play in the dirt. It didn't help that yesterday I was stuck in the house with three kids with stomach flu cleaning chunks out of the carpet....

1/11/2006 10:21:19 AM

Indiana

Vince,
I love your work. I am still deciding how valid it is, but dude, thats awesome that you took the time to do it! Where would a 643 pounder that was 2.91 standard deviations above the mean end up....1 offspring

3 fruit weighing 570, 538, and 490.5 grown in slightly better condition where the same seed produced 1029, 963, and 838 in a competition setting...using this as a control

Its been to long since i had stats to make my brain think that hard

Tom

1/11/2006 12:27:21 PM

Sioux Falls, SD garyboer@dakotalink.com

. . .you weren't born when I took stats. . .

1/11/2006 12:51:02 PM

Menomonie, WI (mail@gr8pumpkin.net)

The thing that bothers me is we are only looking at the top side of the bell curve, and using only the data from the seeds planted that met or exceded expectations. It is very Biased..Although I think your stats skills are excellent. Now if I could only learn the difference between Binomial and Poisson and what applications to use them for..

1/11/2006 3:01:26 PM

Western NY

i wonder how many data points of 0 pounds there are? the pain in the butt ones that aborted everything, or shattered them at softball size.

1/11/2006 3:13:52 PM

SF Bay Area, California

I looked at several ways to contend with all the no-shows and the smaller fruit that are not present in the data. I think the best way to deal with this obvious bias is to put a seed line's results into two seperate buckets. The "dud seed" and the "good seed".

One bucket is for dud seeds - the one's that fail to germinate, are mutants, flat vine and are culled, abort all fruit, killed in the line of duty by insects or vermin, are not grown competitively or produce fruit under 500. The percentage of dud seeds for me has been about 30%. This percentage is grower specific. Other growers may have a better or worse average of getting a plant to the finish line.

The other bucket is for the good seeds in a seed line that sucessfully produce a competitive 500+ pumpkin. The data that is available for this population is fairly complete and that data is what I used for my little stats exercise. My results then should be spot-on for all the "good seeds" in the seed line.

The way to think about it when you are starting the season with a 1068 Wallace seed in your hand is: There is a 30% chance that this seed will end up a dud but if I manage to get this baby to the finish line there's a 57% chance that the fruit will be over 1,000 pounds.

Each grower then can compute his own "dud rate" and use it to compute his own personal odds of success.

vince

1/11/2006 4:40:13 PM

Norton, MA

2002=8 lb. EST from local farm AG seed
2003=370.6 lb. from 869.5 Calai, 1 bad plant, 1 culled seedling, 1 no germ.
2004=63 lb. 1023 Vader disease killed plant, 123 lb. bomb + 516 Aug 1 (approx.) pol. on 481 Hester, 281 on 503 Beauchemin, 8 top seeds steamed at 225* F, 3 good seeds lost to too little heat, 2 no care patch plants no fruit
2005=311* + 351* on no care June 1 planting 771*, 557 on 1104.5, 720.6 on 1068, NO FAILED SEEDS/PLANTS!

Overall out of 24 seeds there were 20 "duds". It wasn't the seeds fault. The best way yet of looking at data for most purposes though. AleX Noel.

1/11/2006 6:30:07 PM

Norton, MA

21 duds, AleX Noel.

1/11/2006 8:27:10 PM

Ohio

Man! I have it easy, All i have to do is count inmates every two hours. I think today Im gonna try to count them by 2's,,,,,,,,,,ok maybe not, but i could if I wanted too, Yea huh, I could.

1/12/2006 4:24:40 AM

Ohio

Im serious, I really could.

1/12/2006 4:27:23 AM