Great! A research paper I read gave me the final formula for some observables and some parameters assuming a gaussian but didn't tell anything about how we actually got there. Instead of searching through books, I stumbled upon this piece of gem and it saved so much of my time. Thank you!
It was really helpful to clear some basic likelihood and Fisher matrix concepts which are used to constrain cosmological parameters... Thanks a lot sir! :)
Thanks very much, a really useful intro. I was coming across references to Fisher matrices in the literature without knowing what they were. Your video means I now have some idea what they are talking about at least. I have ordered the book you mention.
The example at the very end is important for understanding the Fischer matrix formula at the beginning (when "d^2 likelikhood / d lambda^2" is "converted" into "d P/d lambda_alpha dP/dlambda_beta"). This involves a few assumptions that are only explicited at the end…
when at 3:50 he mentions parabola is not a good approximation instead in log space it is gaussian and thats a good approximation .....what does he mean?
i was actually searching a video to explain me fisher matrix since i'm studying cosmology and i found the explanation of the exact piece of book i was studying hahahahahahaha
I am so confused about Likelihood. Isn't likelihood defined as P(theory | data) instead, as described in the wiki link below? But I understand why it can be P(data | theory) as well. en.wikipedia.org/wiki/Likelihood_function#Definition
P(theory|data) is the posterior, which you really want to know (constrain theory from data). To get there if you take Bayesian approach, you need to multiply prior to likelihood, and with uniform prior the likelihood is just posterior up to normalization.
Great! A research paper I read gave me the final formula for some observables and some parameters assuming a gaussian but didn't tell anything about how we actually got there. Instead of searching through books, I stumbled upon this piece of gem and it saved so much of my time. Thank you!
@5:24, the covariance matrix is not in general equal to the inverse of the fisher information matrix. Equality holds only for the Gaussian case.
is it equal to the asymptotic covariance matrix in general?
It was really helpful to clear some basic likelihood and Fisher matrix concepts which are used to constrain cosmological parameters... Thanks a lot sir! :)
i'm a fucking physics graduate in cosmology and this was helpful! - thank you very much :), such a great use of the internet
Thanks very much, a really useful intro. I was coming across references to Fisher matrices in the literature without knowing what they were. Your video means I now have some idea what they are talking about at least. I have ordered the book you mention.
Great Video, where do i find more videos like these explaining concepts in cosmology?
Thank you, I've been studying and this helped tons.
The example at the very end is important for understanding the Fischer matrix formula at the beginning (when "d^2 likelikhood / d lambda^2" is "converted" into "d P/d lambda_alpha dP/dlambda_beta"). This involves a few assumptions that are only explicited at the end…
when at 3:50 he mentions parabola is not a good approximation instead in log space it is gaussian and thats a good approximation .....what does he mean?
Oh man! Thanks a lot for such guidance. It is really valuable.
Very impressive explanation. You have some real teaching skill
What's that word at 7:17? Sound like fedooshal
Many thanks, very much appreciated!
Came here to do my homework about Fisher's information. Stayed for the lesson on cosmology.
great video :D
i was actually searching a video to explain me fisher matrix since i'm studying cosmology and i found the explanation of the exact piece of book i was studying hahahahahahaha
Seth Godin teaching Stats? :)
Thanks
I am so confused about Likelihood. Isn't likelihood defined as P(theory | data) instead, as described in the wiki link below? But I understand why it can be P(data | theory) as well.
en.wikipedia.org/wiki/Likelihood_function#Definition
P(theory|data) is the posterior, which you really want to know (constrain theory from data). To get there if you take Bayesian approach, you need to multiply prior to likelihood, and with uniform prior the likelihood is just posterior up to normalization.
eu sou Brasileiro ñ entenddo sua fala
i know i wanted to watch the hiroshima one cant understand him talk