/sci/ - Science & Math » Thread #12876783

78KiB, 1344x576, cyclic.png

View Same Google iqdb SauceNAO

Econometrics for dummies

Anonymous Wed 24 Mar 13:59:38 2021 No.12876783 View Reply Original Report

Quoted By: >>12876815 >>12877043

How do you fix cyclic time series data with a frequency of 1 year? Bonus points if the advice has Eviews in mind.
(Feel free to call me retarded)

Anonymous

Anonymous Wed 24 Mar 2021 14:16:07 No.12876815 Report

Quoted By: >>12876830

>>12876783
wut?

Anonymous

Anonymous Wed 24 Mar 2021 14:21:42 No.12876830 Report

Quoted By: >>12876960

>>12876815
How do you adjust your data to it's cycles.
Everywhere i've looked everyone just talks about deseasonalizing and just briefly mentions cycles.
Nobody actually talks about even a basic form of the process.

Anonymous

Anonymous Wed 24 Mar 2021 15:14:08 No.12876960 Report

Quoted By: >>12876977

>>12876830
cycles = seasoning

Anonymous

Anonymous Wed 24 Mar 2021 15:19:02 No.12876977 Report

Quoted By: >>12876988

>>12876960
But what if your data are given per year and not per quarter or month? Wouldn't deseasonalizing be impossible then?

Anonymous

Anonymous Wed 24 Mar 2021 15:22:16 No.12876988 Report

Quoted By: >>12877001

>>12876977
How can you see monthly or quarterly data if all you have are yearly data?

Anonymous

Anonymous Wed 24 Mar 2021 15:25:49 No.12877001 Report

Quoted By: >>12877021 >>12877173

>>12876988
You don't, which is why you can't deseasonalize.
Yearly data, instead of seasons, have cycles.
And here comes the trouble, how do you determine those cycles and how do you adjust your data for them?

Anonymous

Anonymous Wed 24 Mar 2021 15:29:43 No.12877021 Report

Quoted By: >>12877074

>>12877001
You can. Those "cycles" are nothing more than long ass seasons.

Anonymous

Anonymous Wed 24 Mar 2021 15:33:23 No.12877043 Report

Quoted By: >>12877080

>>12876783
Econometrics = obsolete time series methods for boomers

Anonymous

Anonymous Wed 24 Mar 2021 15:40:17 No.12877074 Report

Quoted By:

>>12877021
So... same process but different name then?
Might as well try it.

Anonymous

Anonymous Wed 24 Mar 2021 15:41:51 No.12877080 Report

Quoted By: >>12877110

>>12877043
What is the modern version?

Anonymous

Anonymous Wed 24 Mar 2021 15:52:53 No.12877110 Report

Quoted By: >>12877122 >>12877173

>>12877080
Generalized additive models, bayesian structural time series, and deep learning are some more up to date methods.

t. professional data scientist specializing in time series

Anonymous

Anonymous Wed 24 Mar 2021 15:56:01 No.12877122 Report

Quoted By:

>>12877110
Thanks, will make sure to look those up.

Anonymous

Anonymous Wed 24 Mar 2021 16:13:06 No.12877173 Report

Quoted By: >>12877244 >>12878613

>>12877001
Why would you need to "adjust" data for cycles lasting more than a year? Do you want to separate cycles from the trend? You can use a filter like Kalman or Hodrick-Prescott for that.
>>12877110
All of this is econometrics.

Anonymous

Anonymous Wed 24 Mar 2021 16:36:41 No.12877244 Report

Quoted By: >>12877962

>>12877173
If i wanted to create a regression out of that data wouldn't i need to adjust it in order to account for the noise created by the cycle spikes?
Or is that something we do only in quarterly/monthly data and forget about in yearly data?

Anonymous

Anonymous Wed 24 Mar 2021 19:42:03 No.12877962 Report

Quoted By:

>>12877244
I think you first need to understand what you want to "fix" and why. You can't just follow a cookbook not knowing what you're doing and why. An easy fix would be adding a time trend or time dummies in the regression to isolate the time effects but that's not the only thing you need to worry about. Do you have omitted or endogenous variables? Does your data have structural breaks? How are you dealing with that? How confident are you in your assumptions on the statistical properties of the data? Given your model and your data, are you using the right tools for the job? And so on.

Anonymous

Anonymous Wed 24 Mar 2021 22:46:27 No.12878613 Report

Quoted By: >>12878906

>>12877173
>All of this is econometrics.
Econometrics is a subset of time series analysis as applied to economics. You are a retard.

Anonymous

Anonymous Thu 25 Mar 2021 00:12:46 No.12878906 Report

Quoted By:

>>12878613
>econometrics is obsolete and time series are better
>it's the same thing
>no it's the same thing retard
???

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches