Daniel Marcelino dmarcelinobr

💭

I may be slow to respond

Data-driven journalist at @JOTAJornalismo who loves algos, predictive analytics and computational journalism. I do things in R/Python/SQL

dmarcelinobr / forecast5.R

Last active August 29, 2015 14:07

	> summary(mcmc)
	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 5.
	Predicted at 2014-10-05, using 50 polls from 5 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 46 minutes.


	Expected results + swing voters:

dmarcelinobr / valid.votes.R

Last active August 29, 2015 14:07

	> print(wtd.polls)
	PT PSDB PSB OTHERS WASTE n SWING
	VoxPopuli 0.3800000 0.1700000 0.2540000 0.02000000 0.07000000 2000 0.1060
	Sensus 0.3510000 0.2070000 0.2500000 0.02500000 0.08350000 2000 0.0835
	MDA 0.4040000 0.1980000 0.2520000 0.02300000 0.05900000 2002 0.0640
	Datafolha 0.4000000 0.2000000 0.2500000 0.03000000 0.10000000 7520 0.0200
	5 0.4103400 0.2063354 0.2637111 0.02811200 0.09150146 13522 0.0000
	6 0.4516684 0.2271169 0.2902713 0.03094336 0.00000000 13522 0.0000

dmarcelinobr / forecast4.R

Created October 1, 2014 16:40

	> summary(mcmc)
	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 5.
	Predicted at 2014-10-04, using 89 polls from 9 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 45 minutes.


	Expected results + swing voters:

dmarcelinobr / forecast3.R

Created September 27, 2014 14:34

	> summary(mcmc2014)
	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 5.
	Predicted at 2014-09-27, using 37 polls from 5 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 45 minutes.


	Expected results + swing voters:

dmarcelinobr / forecast2.R

Created September 25, 2014 18:35

	R> summary(mcmc2014)
	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 10.
	Predicted at 2014-09-25, using 36 polls from 5 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 57 minutes.


	Expected results + swing voters:

dmarcelinobr / forecast2.R

Last active August 29, 2015 14:06

	R> summary(mcmc2014)
	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 5.
	Predicted at 2014-09-20, using 33 polls from 5 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 28 minutes.


	Expected results + swing voters:

dmarcelinobr / Scotland2.R

Last active August 29, 2015 14:06

dmarcelinobr / forecast1.R

Last active August 29, 2015 14:06

	forecast class object:
	MCMC: 1000 samples with a burnin of 500 and a thinning of 5.
	Predicted at 2014-09-20, using 79 polls from 9 houses.
	Evidence starting at 2014-02-19.
	Election day 2014-10-05.
	Machine runtime: ~ 42 minutes.


	Expected results + swing voters:
	Expected results: (Pooled) National samples only

dmarcelinobr / missing.correction.R

Created September 16, 2014 00:54

	doInstall <- TRUE
	require(Amelia)
	require(ggplot2)

	ANES <- read.csv("ANES.csv")
	ANES <- ANES[ANES$year == 2008, -c(1, 11, 17)] # Limit to just 2008 respondents,
	head(ANES) # remove some non-helpful variables

	with(ANES, plot(jitter(pid7), jitter(ideo7)))
	myModel <- lm(pid7 ~ ideo7 + female + age + south, data = ANES)

dmarcelinobr / prob_AvsB_0.R

Last active August 29, 2015 14:06

Resolve por aproximação o tal do empate técnico estatístico da margem de erro

	############## ##################
	# Qual é a probabilidade de A estar a frente de B dada a diferença entre eles e o erro amostral?
	# Abaixo, a diferença é de .505 - .495 = 1%

	A = .505
	B = .495
	n = c(2500, 1111, 625, 400)

	# Resolve a margem de erro por N
	(moe = sqrt((A(1-A)/n)+(B(1-B)/n)+(2((AB)/n)) ) )