Originally Posted by 2ndMoparMan
1 down, 2 to go.
1 down, 2 to go.
acket.
Yeah, punitive is coming. Hope they take him for everything.
acket.
That's good investigative work by the court and the lawyers, but the court's bias is clearly showing with this verdict.
How can we know for sure that the Sandy Hook massacre wasn't staged?
My concern with the massacre isn't purely because of my distrust of the NRA, but the US really shouldn't also be funding court cases that the plaintiff could be funding themselves.
If it's not staged, then the NRA are more culpable allowing such a dangerous market.
Either way, it's lose lose for the NRA, that's why the NRA is not cooperating. If it's staged, then the US may be more culpable. Or there could be a more sinister plot, that somebody from InforWars intentionally leaked his emails to the plaintiffs.
Anyway, I think only Jesus will know the answer. Bottom line is true Christians would not shoot anyone or go on mass shooting sprees
Shut up, Mick.
acket.
For sure...
It can depend on the context, but generally both statistical and computational are considered mathematical models. Its a square/rectangle kind of thing.
Computational model typically means a complex model with tuneable parameters and lots of data. Its not considered a simple mathematical model (eg weather forecast vs object falling due to gravity). The system may actually be deterministic, but due to being complex or chaotic or insufficient data, it ends up being a prediction. Think weather forecast, modelling of aerodynamics, modelling of complex structures etc.
Unless the system is quite chaotic you can assume more data, more crunching will give a better result.
Now a statistical model is very similar, both require some model/algorithm to make a prediction based on the data.
The difference between statistical model and computational model is more that a statistical model is typically about a system that is not deterministic.. eg humans. So even with better data you might not actually get a better result - especially if you are using past data to make assumptions about future data. That past data could be previous elections, could be votes already counted!
Overall the system depends as much or more of the quality of the assumptions than the amount of data. Especially for elections "the system" is the people and something like Trump can result in people acting quite different to previous elections.
So computational vs statistical is more about the system being modelled than about how they are modelled!
Just to throw in one more thing: Machine Learning (AI)
Machine learning is when there isn't a human built model, but the computer deduces its own model based on the data. Feed it previous election data and results and let it work out patterns. This would also be more helpful if the data was time stamped data, not just the final data. Since you want the model to work as data flows in. So part of the pattern is how the data comes in over time.
The advantage to machine learning is that it could see patterns that are not obvious to humans. The downside is that these patterns might not actually represent causation - they could just be illusions!
These should be more accurate, but would cost a lot more in development and running costs. This style is also quite new. And it is not very transparent.
It can be used to replace both statistical and computational, it doesn't care about the nature of the system, just the data.
Statistical models are easier to build (you can make very simple ones), easier to understand.
Im not sure if anyone has used machine learning on election data.
My system actually had 5 different models that worked on different data. Plus a manual override that is technically a 6th.
The first model worked on data from other areas, eg general swings. This could then be used on an area with few or even no votes counted. It was the least accurate, but worked off the least data. Worked well when timezones meant there could be 3 hours between the start of counting between East and West coast. This would also be useful for the US, Canada and other places with big timezone difference. TV produces didn't like it as showing a predicted result with no votes freaked them. But they did like that the overall totals would go up quickly! So great for "big view", bad for "close up"
Other models used primary vote counts, another used primary vote counts on polling booths compared to prev counts at the same booths.
And since we had preferential (instant runoff) voting there was other models that used preferential counts. The preferential data is the most accurate but also the last to be counted. But it also ran the risk of being completely wrong and needing to be ignored (the electoral commission needed to guess the top 2 candidates in advance, probably got it right 95% of the time.) That issue could be autodetected by the system or manually selected.
So my overall system would slide between models as they got more data, actually giving a weighted average of the models. If the best model had enough data it could be 100% of the final prediction.
Also by having several models the system could also highlight anywhere the models disagreed (either on the winner or if the results differed by more than a preset amount). This could help flag issues. But it also gave us good talking points - something the on-air talent liked. "oh its interesting that our computer is predicting a win for X although they are currently behind on the count - that's because the polling booths counted traditionally go to Y/postal votes aren't included and go to X". So my system was also helping them find things to talk about. It was designed purely to be an on-the-night system.
But this sort of thing is what has been seen with some of these US counts where the leading candidate swaps - my system might have predicted a winner even if they were behind at some points.
These were all human (yes Im human contrary to rumour) designed models - no machine learning. The tuning of the system was actually done by hand - I had some tools, but basically I crunched numbers and then set the parameters. Mostly these were set the same for most seats, with a few seats given extra effort to have an individual tune (due to quirkiness and being close seats, no point putting effort into a safe seat - I had limited time/budget). And generally they were set purely from numbers. It would take a lot of convincing to make my tweak numbers based on actual politics! Amazing how often the tv network would get a political expert to help me and in the end they gave me nothing that improved the models or the tuning.
I didn't need to know the politics, just the numbers!
And my system ran on a single PC computer. It was networked to others, so people had access to the system. But the core data input and crunching was a few programs running on a single server.
I hope that answers your question, sorry (not sorry) about the novel!
Alex Jones has fewer redeeming qualities than Rush Limbaugh did. In case anyone thought that wasn’t possible.
Shut up Billi!That's good investigative work by the court and the lawyers, but the court's bias is clearly showing with this verdict.
How can we know for sure that the Sandy Hook massacre wasn't staged?
My concern with the massacre isn't purely because of my distrust of the NRA, but the US really shouldn't also be funding court cases that the plaintiff could be funding themselves.
If it's not staged, then the NRA are more culpable allowing such a dangerous market.
Either way, it's lose lose for the NRA, that's why the NRA is not cooperating. If it's staged, then the US may be more culpable. Or there could be a more sinister plot, that somebody from InforWars intentionally leaked his emails to the plaintiffs.
Anyway, I think only Jesus will know the answer. Bottom line is true Christians would not shoot anyone or go on mass shooting sprees
But you’re darn right about real Christian shouldn’t shoot people like that. Those who live by the guns die by the guns!
Posting Permissions
Reply With Quote