My main issue is that when analyses and conclusions are not watertight and carefully expressed, you set yourself up for attack. This is particularly true for data that uses correlations to demonstrate associations. I know, I do it for a living. My research is constantly criticised by armchair ‘scientists’ because “correlation is not causation”. And yet I alway clearly point this out in my publications, and I never claim causal links, and I always suggest other explanations, and I never assume my interpretation is the only correct one. Of course, most armchair critics don’t actually read the papers, they just respond to a headline. Funnily, fellow scientists rarely criticise research on these grounds.Oh, and for those who are keen to learn, you might like to read up on the Bradford Hill criteria of causation.

And for those who think r=0.6 is small, it isn’t. It indicates that 36% of the variance in the association between suicide and unemployment is accounted for, in this data-set. That’s not a trivial figure. And no, that doesn’t mean 36% of suicides are due to unemployment, another easy trap to fall into. There are other statistics you can use to calculate the amount of suicide we might be able to attribute to unemployment.

I never claimed unemployment was the sole cause of suicide, read the post. I didn’t so a multiple regression because the point was simply to show that there is a relationship. I don’t have the raw data on suicides, so I couldn’t control for whether suiciders were in the workforce or not. It would stregnthen the data but not having it doesn’t invalidate it.

Look, it’s not a sociology exam, Paul, it’s a blog and we’re already beyond most people’s maths when we say ‘correlation’.

Can you imagine if I had to try to make a quick and nasty explanation of multiple regression at the same time? Half the righties would claim I was just playing tricks to make the numbers say what I want while you and Lew would still be nitpicking.

As I have already said

“It could be that unemployment causes depression and thus suicide, it also could be that depression causes unemployment, it could also be that a third factor causes both or it could also be that the correlation is completely spurious.”

What I was trying to point out to you, and now appear to have succeeded, is that the relationship between unemployment and suicide is not as simple as you originally claimed.

If it is a “multitude of factors” then why use simple regression? If there are a “multitude of factors” then at the very least multiple regression should be used to deal with the other factors than can effect the suicide rate. This brings me back to my original point, that was that your analysis, based as it is on simple regression and a correlation coefficient, doesn’t show the result you are claiming.

As to my discussion about age brackets not making much sense to you, I can believe that. This is just another way I tried to point out to you that the relationship between unemployment and suicide is due to a “multitude of factors”. The point is very simple. If you have a whole lot of people in your sample, some of whom will die due to suicide, but who cannot be unemployed, then it is not clear how unemployment could be a factor in suicide for that group. So one of the “multitude of factors” that should be controlled for is being in the work force and thus being able to be unemployed. If you are in school, for example, and attempt suicide, unemployment is unlikely to be one of the causes. So to get the true marginal effect of unemployment on suicide you may want to remove such people from your sample.

Also I said before what do the references in google search have to do with your “analysis’?

“And the fact is you can’t handle what they show because it conflicts with your ideology.”

What ideology?

@Armchair Critic;
Sorry about that.

they know who the enemy is and, while they don’t refrain from criticising the left, they don’t make an occupation of stabbing their allies in the back.

The ‘enemy’ for me isn’t just ‘those bad righties’ – it’s ‘people who misuse figures and draw spurious conclusions’ as well, regardless of how good their politics might be. I don’t give people a pass just because I agree with them, and if you want to maintain or improve the standard of thought and analysis within your political movement, neither should you.

L

“And here’s the Magg paper;
http://www.kof.ethz.ch/publications/science/pdf/wp_207.pdf”

Let me note this statement from the conclusion to the paper:

“In general however, identified dependence of suicide rates on economic conditions is weak. Typically, significant effects are only present in some age-sex groups, while over-all societal suicide rates are unaffected by economic conditions. This result is consistent with Rodriguez-Andres (2005) and Gerdtham and Ruhm (2006). Both studies report that economic variables are insignificant once country fixed effects are accounted for.”

L

1- your staunch defence of the sell-outs in the Maori Party because you believe in the ideal they’re meant to represent (fair enough) but can’t face up to the fact that they have sold out, and

I think it’s too early to call them as sell-outs, and I will happily do so if they prove unable to make any substantive gains in exchange for their support of this government. But remember, I’m not so much concerned with limited tactical policy wins as with strategic symbolic gains or gains in political infrastructure. They’re already well on the way to making just such a gain with the Foreshore and Seabed Act review, about which I think we’re due to hear back in the next week or so. I will be very concerned if they don’t take a stronger stance on the matter of mana whenua seats than what they have done so far. And none of this ‘non-voting representation’ stuff, either; that’d be inadequate.

However I’m patient, and I don’t necessarily think they should quit the government over the issue.

so you don’t like the Standard for putting the fact they’re sell-outs in your face

It’s not a fact.

To throw a bit of counter-motive speculation back at you: a lot of Labour partisans around here want the māori party to fail and for Māori to come crawling back to Labour. That’s the thing which pisses me off most – people calling them done before they’ve even gotten going.

2- the relative success of your own blog’s approach which is aloof, intellectual, and hectoring compared to the earthier battler style of this blog

Well, for one thing it’s not my blog; I’m the second-to-most-recent poster to join and it was already well underway when I did. For another thing, it’s doing just what we want it to – the ‘earthiness’ of which you speak isn’t what we’re pitching for and we take efforts to discourage quantity in favour of quality. YMMV; if you don’t like it, kindly don’t bother with it.

I’m certainly not envious – I turned down an invitation to be a poster here in mid-2008.

L

Where did anyone say that unemployment is a sole cause of suicide? i have argued that higher unemployment increases the numbers of suicides – it doesn’t have to be a sole cause in a single case for that to happen.

Still, at least now you’ve admitted that this the basis of your complaint is you don’t want to accept the conclusion.

I’ll try to explain it for you. Every human has a propensity to any given behaviour, whether that propensity is realised will depend on a multitude of factors.

You like reducing human behaviour to numbers so I’ll give you an analogy. Say everyone has a suicide score, when it hits 100, they kill themselves. Everyone will have different base scores and different added points due to events or circumstances in their lives. Let’s say unemployment is worth 10 points. It’s not in and of itself going to be a sole cause for anyone but for those who already have other pressures, it could push them over the line. So, with a general increase in unemployment you’re going to get some more people killing themselves.

Your ramblings about age brackets don’t make much sense but here’s how i got the numbers – the suicide numbers are straight from the report – the youth bracket in that report is 15-24. The unemployment numbers are the number of unemployed from the age groups 15-18 and 18-24 combined divided by the labour force (note labour force, not population) for those age groups. The trend line and it’s slope are determined by excel. The correlation equation also. I haven’t done anything to the numbers, its a straight up and down report of what they show.

And the fact is you can’t handle what they show because it conflicts with your ideology.

Marty. For a start none of the references in that search have anything to do with your “analysis”. Your regressions and your correlation shows causation notion still has problems. For example, I’m told that a youth in your data set is someone 15-24, so what’s with the increase in the slope of your youth regression? For 15-18 group unemployment doesn’t look like much of an issue since they are in school. Add to that the over 18s who are still in some form of education and thus unemployment isn’t an issue and I don’t see where the increase in the slope coefficient is coming from. It has to be down to those unemployed in the 18-24 group. It seems a big jump for just this group. The suicide rate could increase for the 15-24 group but for it to be due to just unemployment seems unlikely given the number in that group who are not in the labour force. Your “analysis” should have accounted for these factors.

To take just one example from your google search:

“Relationship between the Economy, Unemployment and Suicide”
Prepared by the Suicide Prevention Resource Center
November 12, 2008

It states

“Unemployment contributes to suicide risk, but does not “cause’ suicides on its own.

* Employment status is but one of dozens of factors that interact dynamically within individuals, communities, and societies and affect the risk for suicide.

* Although unemployment is associated with increased rates of suicide, many individuals may have lower rates of employment because of mental health and/or substance abuse problems, which are also associated with increased suicide rates.”

Notice that the relationship between unemployment and suicide is much more dynamic than your “analysis” would suggest. Also note the second point. Unemployment may be “caused ” by mental health issues that are associated with increased suicide rates. As I tried to point out, a correlation between unemployment and suicide doesn’t just mean “unemployment causes suicide”, the interactions between the two can run in both directions or be caused by a third factor.

Also the report says

“We can expect a sharp downturn in the economy to increase suicide risk, especially among working-age adults and older adults whose retirement security is threatened.”

Why then do you show the increase in suicide rate for the youth group? Your “analysis” seems to be missing something.

Paul’s more pointing out some rather basic issues with correlation, though he should have pointed out more clearly that a empirically testable and/or valid mechanism is needed to explain the correlation.

I’d also disagree that the 0.40-0.60 correlation’s are “substandard”, but then I’m a biology undergrad and used to the messiness of biological data. Where correlation’s of similar strength thrown up in the initial analysis are useful for indicating whether or not there’s a relationship between two (or more) variables worth digging into the more esoteric forms of data analysis for, and indicating research directions.

Though briefly looking at the paper gomango mentions, (and fails to bloody link to) it appears that there’s no significant predictive power for unemployment levels on suicide rates across the OECD. But, as Paul sort of mentioned, other factors can come into play that may confound analyses, and need to be debugged, i.e. cultural factors may come more into play, though not having the time, nor the concentration (ADHD + sleep deprivation), and the econ/stats backgrounds I can’t read deeper into this…

And here’s the Magg paper;
http://www.kof.ethz.ch/publications/science/pdf/wp_207.pdf

as for my suggestion that unemployment increases suicide http://www.google.com/search?q=suicide+unemployment&rls=com.microsoft:en-nz:IE-SearchBox&ie=UTF-8&oe=UTF-8&sourceid=ie7&rlz=1I7GGLL_en

and do some damn reading http://www.google.com/search?q=suicide+unemployment&rls=com.microsoft:en-nz:IE-SearchBox&ie=UTF-8&oe=UTF-8&sourceid=ie7&rlz=1I7GGLL_en