1
Reposts
3
Likes
13
Replies
(Edited)
I really like the notion of decomposing anything into more intuitive or structurally understood components, the entire purpose of many matrix decompositions is to observe structure. You are correct in that spot growth today to back out or decompose informations offers a causality issue in the time series sense more specifically this turns into a “smoothing” problem where you have some state (or states) you are decomposing and couldn’t directly observe past tense so you are trying to observe them in the past now, likely more appropriate to use contemporaneous spot values so it’s apples to apples in the time sense. I like the idea and would be eager to see you share a small code snippet and some pictures trying to implement it on a small scale for further discussion. Thanks for sharing! (Also! I see you have equations, use latex in your post to render the math!)
Reply
Loading replies...