***2024 murder toll thread***

[pugboy]

expecting heinz to go in his constituency to have a walk about and plead with the youths to stop killing each other but can keep on doing home invasions

[Dohplaydat]

unless a miracle, there will be more than 600 this year
not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

[zoom rader]

expecting heinz to go in his constituency to have a walk about and plead with the youths to stop killing each other but can keep on doing home invasions

Yeah he tells them to pick up a book

[pugboy]

have they started singing as yet?

expecting heinz to go in his constituency to have a walk about and plead with the youths to stop killing each other but can keep on doing home invasions

Yeah he tells them to pick up a book

[pugboy]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

unless a miracle, there will be more than 600 this year
not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

[The_Honourable]

374: Garnett Sydney shot and killed in rio claro

[The_Honourable]

361: rohan dindial was found dead st his enterprise home on July 29th. Autopsy results reveal he died from blunt force trauma

So far police haven't classified this as a murder (yet... although "kinda obvious"), you think we should drop the toll by 1 until confirmation?

[shake d livin wake d dead]

Yea drop it before habit start to cry^... So that's 373

But an unidentified male was shot and killed in Mon repo road morvant.... 374

[pugboy]

like the road blocks causing more murders
where the joint patrols rowlee keep talking about

[Dohplaydat]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

unless a miracle, there will be more than 600 this year


not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

Ideally we'd want to have the number of murders per day for the last few years and use a time series forecasting model like Prophet from Meta to predict this. That model would consider seasonal trends and spike prediction. It's what I use for my stock portfolio and it's quite accurate but can get complicated.

[shake d livin wake d dead]

like the road blocks causing more murders
where the joint patrols rowlee keep talking about

Joint patrols between license office and ttps in full swing

[FrankChag]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

unless a miracle, there will be more than 600 this year


not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

Ideally we'd want to have the number of murders per day for the last few years and use a time series forecasting model like Prophet from Meta to predict this. That model would consider seasonal trends and spike prediction. It's what I use for my stock portfolio and it's quite accurate but can get complicated.

@pugboy A graph in the 2024 Murders thread.. hmm... I wonder what the x-axis is, and what those figures are.. or even what 1.12x^2 + 44.7x + 42.9 == 670.12 and 51.4x + 37.3 == 602.7 below the graph means.. sigh. If you really wanted to critique the graph, you would note that it is 2024 data, and the y axis is labeled 2020. Or perhaps you didn't notice that my preceding posts were discussing the data, and they got cut off into the previous page.

As regards to the modelling, well shoot: respect... so there's another stats nerd on here. Lots of them since GATE (although some are posers bc they know big words like 'distribution' and 'model fit'). Some time ago, when we had the discussion on looking at areas, I started pulling together the daily murder rates with the areas from each tread for the 5 years, to do exactly that, an econometric time series analysis, and then some multivariate analyses by area. I was way into scripting the models when I realized that, you know what, this is someone else's job, and I already have years of infuriating public sector experience where I was literally doing other "get-thru" people's jobs, sometimes carrying an entire team. I don't do that anymore.

I mean I can pull out my Masters econometrics and PhD advanced stats exam grades, and all the real stats work I've done I between, but I'd rather not have a "measuring", chest-thumping contests with randos on the internet.

Although, if someone who really knew what they were talking about wanted to critique the approach though, they'd say something like:

-- Regression models on 6 or 7 data points? Seriously?
-- Where are your r-sq and p values?
-- What modelling approach, IVs, ModVs, MedV, etc., have other scholars used?
-- Also, is a time-series really the correct approach to model murders?

I'm not, nor do I want to be, a Research Criminologist, but then you need to ask yourself, after doing all that work, what is the corresponding increase in measurement precision with so many variables... and at the end, was it even worth it?... It's not published work, and the policy makers appear to be corrupt, and internet readers are fickle and often clueless.

So a simple average the linear model and the polynomial model is good enough precision for me, as an arm-chair estimate.


Besides, who really cares?

If you do, then produce your own model, and post your results.

Otherwise, thanks for your.. feedback. I'll take it under advisement.

[pugboy]

frank, you seem upset

[FrankChag]

frank, you seem upset

Hmm.. have you seen the figures?
And the weak response?

Aren't you upset?

July-2024 was the bloodiest month on record.

69 murders.

8 more ie you hit 382, and there would have been 3500 murders since 2018, by tuner counts.

what fun.

[FrankChag]

frank, you seem upset

Also...

[The_Honourable]

Yea drop it before habit start to cry^... So that's 373

But an unidentified male was shot and killed in Mon repo road morvant.... 374

Identified as Ian Charles

[The_Honourable]

375: Anjanee Latchman Isidore stabbed and beaten to death in maraval

[shake d livin wake d dead]

Home invasion victim ^

[Dohplaydat]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

unless a miracle, there will be more than 600 this year


not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

Ideally we'd want to have the number of murders per day for the last few years and use a time series forecasting model like Prophet from Meta to predict this. That model would consider seasonal trends and spike prediction. It's what I use for my stock portfolio and it's quite accurate but can get complicated.

@pugboy A graph in the 2024 Murders thread.. hmm... I wonder what the x-axis is, and what those figures are.. or even what 1.12x^2 + 44.7x + 42.9 == 670.12 and 51.4x + 37.3 == 602.7 below the graph means.. sigh. If you really wanted to critique the graph, you would note that it is 2024 data, and the y axis is labeled 2020. Or perhaps you didn't notice that my preceding posts were discussing the data, and they got cut off into the previous page.

As regards to the modelling, well shoot: respect... so there's another stats nerd on here. Lots of them since GATE (although some are posers bc they know big words like 'distribution' and 'model fit'). Some time ago, when we had the discussion on looking at areas, I started pulling together the daily murder rates with the areas from each tread for the 5 years, to do exactly that, an econometric time series analysis, and then some multivariate analyses by area. I was way into scripting the models when I realized that, you know what, this is someone else's job, and I already have years of infuriating public sector experience where I was literally doing other "get-thru" people's jobs, sometimes carrying an entire team. I don't do that anymore.

I mean I can pull out my Masters econometrics and PhD advanced stats exam grades, and all the real stats work I've done I between, but I'd rather not have a "measuring", chest-thumping contests with randos on the internet.

Although, if someone who really knew what they were talking about wanted to critique the approach though, they'd say something like:

-- Regression models on 6 or 7 data points? Seriously?
-- Where are your r-sq and p values?
-- What modelling approach, IVs, ModVs, MedV, etc., have other scholars used?
-- Also, is a time-series really the correct approach to model murders?

I'm not, nor do I want to be, a Research Criminologist, but then you need to ask yourself, after doing all that work, what is the corresponding increase in measurement precision with so many variables... and at the end, was it even worth it?... It's not published work, and the policy makers appear to be corrupt, and internet readers are fickle and often clueless.

So a simple average the linear model and the polynomial model is good enough precision for me, as an arm-chair estimate.


Besides, who really cares?

If you do, then produce your own model, and post your results.

Otherwise, thanks for your.. feedback. I'll take it under advisement.

Clearly you care, I was trying to help as you seemed to enjoy doing it.

I have done a lot of stats in my lifetime. You can do a regression with 3 points and get good results if the underlying distribution is linear, yet you went poly already, talk about over fitting.
Simple models are best and easiest to explain, I don't think I need to state why it's obvious a poly nominal model is not appropriate here.

That's why I recommended getting the data per day and doing a time series analysis.

Your R sq, mse, Mae, p-values doesn't matter because your model is lacking data to make any of this statistically significant.

Stop behaving like a wrong and strong typical Trini and take constructive criticism.

[Mmoney607]

Home invasion victim ^

Most likely not home invasion.....

[FrankChag]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

unless a miracle, there will be more than 600 this year


not much sense trying to extrapolate anymore

good in a sense that erla and heinz not going to have a number to keep bumping they gums about how they kept below the 600 mark

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

Ideally we'd want to have the number of murders per day for the last few years and use a time series forecasting model like Prophet from Meta to predict this. That model would consider seasonal trends and spike prediction. It's what I use for my stock portfolio and it's quite accurate but can get complicated.

@pugboy A graph in the 2024 Murders thread.. hmm... I wonder what the x-axis is, and what those figures are.. or even what 1.12x^2 + 44.7x + 42.9 == 670.12 and 51.4x + 37.3 == 602.7 below the graph means.. sigh. If you really wanted to critique the graph, you would note that it is 2024 data, and the y axis is labeled 2020. Or perhaps you didn't notice that my preceding posts were discussing the data, and they got cut off into the previous page.

As regards to the modelling, well shoot: respect... so there's another stats nerd on here. Lots of them since GATE (although some are posers bc they know big words like 'distribution' and 'model fit'). Some time ago, when we had the discussion on looking at areas, I started pulling together the daily murder rates with the areas from each tread for the 5 years, to do exactly that, an econometric time series analysis, and then some multivariate analyses by area. I was way into scripting the models when I realized that, you know what, this is someone else's job, and I already have years of infuriating public sector experience where I was literally doing other "get-thru" people's jobs, sometimes carrying an entire team. I don't do that anymore.

I mean I can pull out my Masters econometrics and PhD advanced stats exam grades, and all the real stats work I've done I between, but I'd rather not have a "measuring", chest-thumping contests with randos on the internet.

Although, if someone who really knew what they were talking about wanted to critique the approach though, they'd say something like:

-- Regression models on 6 or 7 data points? Seriously?
-- Where are your r-sq and p values?
-- What modelling approach, IVs, ModVs, MedV, etc., have other scholars used?
-- Also, is a time-series really the correct approach to model murders?

I'm not, nor do I want to be, a Research Criminologist, but then you need to ask yourself, after doing all that work, what is the corresponding increase in measurement precision with so many variables... and at the end, was it even worth it?... It's not published work, and the policy makers appear to be corrupt, and internet readers are fickle and often clueless.

So a simple average the linear model and the polynomial model is good enough precision for me, as an arm-chair estimate.


Besides, who really cares?

If you do, then produce your own model, and post your results.

Otherwise, thanks for your.. feedback. I'll take it under advisement.

Clearly you care, I was trying to help as you seemed to enjoy doing it.

I have done a lot of stats in my lifetime. You can do a regression with 3 points and get good results if the underlying distribution is linear, yet you went poly already, talk about over fitting.
Simple models are best and easiest to explain, I don't think I need to state why it's obvious a poly nominal model is not appropriate here.

That's why I recommended getting the data per day and doing a time series analysis.

Your R sq, mse, Mae, p-values doesn't matter because your model is lacking data to make any of this statistically significant.

Stop behaving like a wrong and strong typical Trini and take constructive criticism.

Because there isn't a linear model there first?
Because I didn't say that I'm averaging the two models as an armchair statistic?

3 data points acceptable because the underlying distribution is "linear"? What BS are you talking?

You were trying to "help me" ??

Serious question: Do you even know how regression works?


Ustedes sigan,, si quieren. Ya no puedo más.

Some people... they read investopedia articles and think they're statisticians.

You go brother.

[alfa]

So we're on track for 600+ murders this year. Can we leave out all the Maths Olympiad geek stuff

[Dohplaydat]

that label less graph doesn’t make sense

i am curious to know how they does “model” murder projections over the year
is it that they use historical patterns?
eg slow numbers carnival time

We're on track for 625+ now

The poly model seems to suggest that, but there was a serious uptick in July.

If it's another 2022 situation, well, then probably.

The worst part is there is no or slow justice...

Cool analysis but these graphs are painful to watch.

No x axis labels, no projection, there's no need to show the formula for the line, and I'd say you're over fitting by trying to use a poly nominal. Perhaps, look at the past few years and model each month. Or interestingly, look at predicting spikes, they probably can be modelled with a Poisson distribution.

Ideally we'd want to have the number of murders per day for the last few years and use a time series forecasting model like Prophet from Meta to predict this. That model would consider seasonal trends and spike prediction. It's what I use for my stock portfolio and it's quite accurate but can get complicated.

@pugboy A graph in the 2024 Murders thread.. hmm... I wonder what the x-axis is, and what those figures are.. or even what 1.12x^2 + 44.7x + 42.9 == 670.12 and 51.4x + 37.3 == 602.7 below the graph means.. sigh. If you really wanted to critique the graph, you would note that it is 2024 data, and the y axis is labeled 2020. Or perhaps you didn't notice that my preceding posts were discussing the data, and they got cut off into the previous page.

As regards to the modelling, well shoot: respect... so there's another stats nerd on here. Lots of them since GATE (although some are posers bc they know big words like 'distribution' and 'model fit'). Some time ago, when we had the discussion on looking at areas, I started pulling together the daily murder rates with the areas from each tread for the 5 years, to do exactly that, an econometric time series analysis, and then some multivariate analyses by area. I was way into scripting the models when I realized that, you know what, this is someone else's job, and I already have years of infuriating public sector experience where I was literally doing other "get-thru" people's jobs, sometimes carrying an entire team. I don't do that anymore.

I mean I can pull out my Masters econometrics and PhD advanced stats exam grades, and all the real stats work I've done I between, but I'd rather not have a "measuring", chest-thumping contests with randos on the internet.

Although, if someone who really knew what they were talking about wanted to critique the approach though, they'd say something like:

-- Regression models on 6 or 7 data points? Seriously?
-- Where are your r-sq and p values?
-- What modelling approach, IVs, ModVs, MedV, etc., have other scholars used?
-- Also, is a time-series really the correct approach to model murders?

I'm not, nor do I want to be, a Research Criminologist, but then you need to ask yourself, after doing all that work, what is the corresponding increase in measurement precision with so many variables... and at the end, was it even worth it?... It's not published work, and the policy makers appear to be corrupt, and internet readers are fickle and often clueless.

So a simple average the linear model and the polynomial model is good enough precision for me, as an arm-chair estimate.


Besides, who really cares?

If you do, then produce your own model, and post your results.

Otherwise, thanks for your.. feedback. I'll take it under advisement.

Clearly you care, I was trying to help as you seemed to enjoy doing it.

I have done a lot of stats in my lifetime. You can do a regression with 3 points and get good results if the underlying distribution is linear, yet you went poly already, talk about over fitting.
Simple models are best and easiest to explain, I don't think I need to state why it's obvious a poly nominal model is not appropriate here.

That's why I recommended getting the data per day and doing a time series analysis.

Your R sq, mse, Mae, p-values doesn't matter because your model is lacking data to make any of this statistically significant.

Stop behaving like a wrong and strong typical Trini and take constructive criticism.

Because there isn't a linear model there first?
Because I didn't say that I'm averaging the two models as an armchair statistic?

3 data points acceptable because the underlying distribution is "linear"? What BS are you talking?

You were trying to "help me" ??

Serious question: Do you even know how regression works?


Ustedes sigan,, si quieren. Ya no puedo más.

Some people... they read investopedia articles and think they're statisticians.

You go brother.

Guesstimating is fine bro. You just don't seem to understand the fundamentals, based on your plot and attitude.

[Mmoney607]

Post Allyuh CV

[pugboy]

i have a bachelors in math and compsci but that has little bearing on this

i found the poly stuff strange and well the labelless graph

[mero]

Frank looking to hit men 7.62 for he graph yes

[paid_influencer]

best practice say to make your communication understandable to the person receiving the information

not really to get defensive and flex with jargon. not that there is anything wrong with you doing that it is helpful to somebody

what i want to know is if erla doing a better job than gary that is what i want to know from the graph is there a decrease in the increase

[16 cycles]

TTPS, CoP or MoNS not coming here to analyze stats....and none of those stats offer any solution though it may help to provide some...

[pugboy]

habit says erla doing a better job and home invasions down

best practice say to make your communication understandable to the person receiving the information

not really to get defensive and flex with jargon. not that there is anything wrong with you doing that it is helpful to somebody

what i want to know is if erla doing a better job than gary that is what i want to know from the graph is there a decrease in the increase

[K74T]

Home invasion victim ^

Most likely not home invasion.....

The Maraval woman who was found lying in a pool of blood on her kitchen floor on Tuesday afternoon, may have been killed as a result of a family dispute.

Residents of Saut D’Eau Road, Maraval, yesterday said while 65-year-old mother of four Anjanee Lutchman-Isidore had a handful of friends, she also had enemies who would have wanted her dead.

The victim’s body was discovered by a relative of her husband’s around 5.36 pm on Tuesday, after she failed to respond to her husband’s calls.

Lutchman-Isidore’s husband is 95 years old and is bedridden as a result of a stroke. She was his primary caregiver.

While no one was at home when Guardian Media visited the area yesterday, residents living close by said the victim used to keep to herself.

But one man said, “She was very friendly to me.”

Saying Lutchman-Isidore moved to the area at least ten years ago after marrying her husband, he said there had been an ongoing dispute over the family-owned property which was nestled on the hillside in Maraval.

Guardian Media was told, “She had her way.”

Another person claimed her presence may have “agitated” other relatives.

Police initially claimed her death may have occurred during a robbery, but it was uncertain if and what was taken from the house up to late yesterday.

When Guardian Media went to the house, several security cameras were seen around the perimeter of the two-storey property, which also boasted several apartments.

The victim’s body was found lying behind a grey metal kitchen door located on the northern side of the compound, with a wound to the head.

Residents said they were not fearful of being victims of crime after learning of the killing, as they strongly believe it is a domestic matter.

An autopsy was scheduled to be performed at the Forensic Science Centre, St James, yesterday.