search

PardeeCenterDU / IFs-Issues-Tracking

This repository only holds the list of bugs that have been reported for IFs. Anyone may add a bug report, but please look to see if your issue has already been added!
2 stars 0 forks source link

First and Second Year of Influence series #85

Closed caleb-petry closed 2 years ago

caleb-petry commented 3 years ago

Currently, there is not a separate procedure for the first two years of the Dependence. Bandwidth, or Influence series. This raises a potential concern. How is the 3-year moving average for the first two years being handled?

An additional but potentially related issue, why is it that a fairly large portion of first-year values are recorded as “.” if we have data in 2016? For context, the United States has four and China has 90 such values in the Influence forecasts.

PardeeCenterIFs commented 3 years ago

I don't see a 3-year moving average for Dependence at all, where are you seeing this logic?

caleb-petry commented 3 years ago

My apologies, this wasn't entirely clear.

The 3-year moving averages are calculated for the individual components rather than for the final series. For example, starting on line 17022:

'Economic Dependence 'Aid % GDP '--------- For k = 2 To 1 Step -1 aid_a_gdp_b(i, j, k + 1) = aid_a_gdp_b(i, j, k) Next k aid_a_gdp_b(i, j, 1) = XDYAD2(i, j, 3) / 1000000000 / SGDP(j) * 100 'jrs 2020/07/10 added / 1000000000 aid_a_gdp_b_ma = 0 For k = 1 To yrMA aid_a_gdp_b_ma = aid_a_gdp_b_ma + aid_a_gdp_b(i, j, k) Next k aid_a_gdp_b_ma = aid_a_gdp_b_ma / yrMA

PardeeCenterIFs commented 3 years ago

On the first issue, in the first year, there's no average, we simply use the forecast value, in the second year, we do a 2 year average, using the first year, and the second.

PardeeCenterIFs commented 3 years ago

Can you show me those 4 "." values for the US in a screen-shot?

caleb-petry commented 3 years ago

Of course, this is coming from a 7.73 version:

image

Moldova alone has values displayed for Dependence or Bandwidth. So, is it correct that the "." simply represents a value so small that it is not displayed?

If so, this is more of an initial-year transient issue. I can dig into it a bit more and report back in that case.

caleb-petry commented 3 years ago

I took a look at it things. The initial-year jumps are likely being caused by something in the other issue threads on here and/or arms trade, for these four specifically.

What is left to address this issue thread and produce better-looking forecasts is to have the first year forecast use two years of historical data and the second year using one. However, implementing this now would make it harder to pick up issues like the initial-year values being incorrect for those four.

To that end, we should add a switch to turn on/off the first two years using historical data versus the current approach if we were to make these proposed changes. Alternatively, we could wait to implement anything here until we have identified any other changes or issues and addressed the following:

To make any progress in addressing the initial-year values, the first two would have to be implemented. The issue relating to Alliances and LoR only matters in later years.