vayuu1 Posted May 17, 2014 Posted May 17, 2014 I don't know how they are getting those figures but seriously ,they are doing something right have been on the spot almost every time,undoubtedly the best exit polls researchers in india.Take a bow,have been on money in 04,in delhi election 2014 and these elections,they deserve a thread and lol at times now -OrG sent from dil se
Crookbond Posted May 17, 2014 Posted May 17, 2014 They have got their numbers right EVERY SINGLE TIME. I posted about them earlier too - State Assembly Elections, LS 2009 etc.
Yoda-esque Posted May 18, 2014 Posted May 18, 2014 TC only had a sampling size of 38K,that's astonishing
Crookbond Posted May 18, 2014 Posted May 18, 2014 TC only had a sampling size of 38K' date='that's astonishing[/quote'] Sample size has got little to do with predictions. That's not the thing you show be astonished about.
ravishingravi Posted May 19, 2014 Posted May 19, 2014 Sample size has got little to do with predictions. That's not the thing you show be astonished about. Why ? Stats are about samples. Their size and quality. If you have doubts on quality or representativeness you increase the size.
Haarkarjeetgaye Posted May 19, 2014 Posted May 19, 2014 its the intelligence you tie with a sample size like what background, religion, caste etc.... Times now gave 249 seats to NDA. what duds
Sachin=GOD Posted May 19, 2014 Posted May 19, 2014 Why ? Stats are about samples. Their size and quality. If you have doubts on quality or representativeness you increase the size. Accuracy of the sample depends more on quality rather than quantity.
Yoda-esque Posted May 19, 2014 Posted May 19, 2014 Accuracy of the sample depends more on quality rather than quantity. Don't state the obvious.I was implying that the Agency has done a great job in getting data points right with demographics as varied as these.
Sachin=GOD Posted May 19, 2014 Posted May 19, 2014 Don't state the obvious.I was implying that the Agency has done a great job in getting data points right with demographics as varied as these. I wasn't replying to you.
Sachin=GOD Posted May 19, 2014 Posted May 19, 2014 Wow!!!! they predicted 0 for Cong in all these states and got all right - remarkable.
Crookbond Posted May 19, 2014 Posted May 19, 2014 Why ? Stats are about samples. Their size and quality. If you have doubts on quality or representativeness you increase the size. No. You can have a large sample size with absolute bizarre representation. Want an example? Population of 100 people - 90 Men, 10 Women. Sample size 50M, 60M, 70M - all are poor representations. Of course, there is a sample size threshold at a confidence interval but beyond that threshold it is not the sample size but "How you pick the size" is the interesting part of the job.
Crookbond Posted May 19, 2014 Posted May 19, 2014 Don't state the obvious.I was implying that the Agency has done a great job in getting data points right with demographics as varied as these. That is again difficult but not exactly rocket science. Most exit poll agencies go for a random sample across wards. The most difficult part is converting vote share to seats where the real science comes into play - non obvious. S=G, Read this article. http://www.newslaundry.com/2014/05/19/todays-chanakya-genius-dumb-luck-or-voodoo/ :facepalm: ICF censors the word "luck" :wtf:
Sachin=GOD Posted May 19, 2014 Posted May 19, 2014 That is again difficult but not exactly rocket science. Most exit poll agencies go for a random sample across wards. The most difficult part is converting vote share to seats where the real science comes into play - non obvious. S=G, Read this article. http://www.newslaundry.com/2014/05/19/todays-chanakya-genius-dumb-luck-or-voodoo/ :facepalm: ICF censors the word "luck" :wtf: read that today. Other exit polls will still be wrong most of the time..
ravishingravi Posted May 19, 2014 Posted May 19, 2014 No. You can have a large sample size with absolute bizarre representation. Want an example? Population of 100 people - 90 Men' date=' 10 Women. Sample size 50M, 60M, 70M - all are poor representations. Of course, there is a sample size threshold at a confidence interval but beyond that threshold it is not the sample size but "How you pick the size" is the interesting part of the job.[/quote'] True. Rule of thumb however is that increasing the size of sample should ensure that it is more representative of population. Your good old t test and z test. I think it is a chicken and egg situation. Chanakya had small sample size cause it was more representative of pop. That is quality of sampling.
Crookbond Posted May 19, 2014 Posted May 19, 2014 True. Rule of thumb however is that increasing the size of sample should ensure that it is more representative of population. Your good old t test and z test. I think it is a chicken and egg situation. Chanakya had small sample size cause it was more representative of pop. That is quality of sampling. No it doesn't, particularly in India when demographics vary. In a homogenous demographic, what you say holds true. However in practice, it is difficult to find homogenous populations. Sample sizes are ALWAYS small.
BlackMamba Posted May 19, 2014 Posted May 19, 2014 No it doesn't, particularly in India when demographics vary. In a homogenous demographic, what you say holds true. However in practice, it is difficult to find homogenous populations. Sample sizes are ALWAYS small. Saying India's demographics vary is an understatement. The breadth and depth of how it varies is quite simply mind-boggling.
Recommended Posts