Reply to: Oh dear: Unemployment at a six year high

[AtW]

The big unknown is number of people on disability benefits - graph of them should have really been added as well just to see if drop in unemployment claims happens to have similar increase in disability claims.

[hyperD]

You say the fact that it's going down is noise, who knows if it's noise or not?

Because of the intrinsic link between total unemployment and unemployment benefit claimants. I would have said the same had total unemployment been going down and unemployment benefits had spiked upwards.

I would have had as much contempt for Cameron saying the latter as I already have for bliar for publically stating the former.

[Numptycorner]

Not my objective. I don't make up the figures.

You say the fact that it's going down is noise, who knows if it's noise or not?

[hyperD]

no, you are guessing it is noise because it suits your objective.

Not my objective. I don't make up the figures.

[AtW]

Have you seen average (for whole hospital) tempreture charts? They were common place in USSR and hospitals were judged on the basis of whether average temp was okay, come to think of it sounds very much NL and NHS

Let's say you've got baseline unemployment that is around 1 mln. One month it deeps down 10k people, which is 1% and you would not even notice it if the chart is done properly - from zero. However if you get it close to your eyes (which is what change of cut off point does), then suddenly that 1% may well be a huge increase - in this case it is used solely to influence weak human brains because objectively 1% change in situation is not a big thing, unless we talk critical values like in case of tempreture (anything below 36C and above 37C for example should be monitored on its own scale)

[wendigo100]

Not using 0 based axis in order to make steepness of change show off better is a sign of a liar using stats to prove his point: should be a fireable offence.

Whoa, hold on a minute. Small deviations would be much harder to view and to analyse. It is perfectly valid to explore only the top of a graph.

A 1% change in unemployment is a very significant event, but you would be pushed to see it on a zero-based graph.

Suppose you applied that principle to a patient's temperature graph in hospital. Life-threatening deviations might go undetected.

[wonderwaif]

That's called a "chickens coming home to roost" curve.

Yes yes yes!, but I remember how bad the Tories were.(I think that is the correct response )

[AtW]

I was looking at the steepness of the Total Unemployment series.

Not starting from zero is legitimate because the series will never get anywhere near it and it would hide information.

Yes and this is why counting should start from zero - unemployment indeed will never reach that level, and this is why steepness of change maybe misleading because if it overall change level is 3-5% then it could be just noise.

Not using 0 based axis in order to make steepness of change show off better is a sign of a liar using stats to prove his point: should be a fireable offence.

[Numptycorner]

For the hard of hearing.... y-axis. They are ALMOST the same scale of magnitude.

The main point of interest is that the rate of change of unemployment is increasing. However, much political capital will be made of the slight dip in benefits claimants which is mainly noise.

no, you are guessing it is noise because it suits your objective.

[CaribbeanPirate]

I was looking at the steepness of the Total Unemployment series.

Not starting from zero is legitimate because the series will never get anywhere near it and it would hide information. Having a common axis is a valid criticism as both series are supposed to be different measures of the same thing i.e. unemployment. However, in reality they have less to do with each other than is claimed or believed.

[hyperD]

no it was a serious question,

the chart has two x axis, one on the right and one on the left and both are scaled differently which makes the lines valueless as the trends are not making a comparison against the same scale

Milan.

For the hard of hearing.... y-axis. They are ALMOST the same scale of magnitude.

The main point of interest is that the rate of change of unemployment is increasing. However, much political capital will be made of the slight dip in benefits claimants which is mainly noise.

Yes, a dodgy graph that all sorts of rabbits will be pulled out of a hat to justify a plethora of failing policies from this incumbent administration.

Certainly a "chickens coming home to roost curve" - like it!

[AtW]

Curves correlate with each other because they show similar thing just calculated differntly to suit whatever absolute figures are best needed.

Unemployment is up but house prices ain't falling, what gives?

[wendigo100]

It compares curves. Note the sudden deviation over the past year.

[AtW]

Well, if I got a report with this chart in it I would be asking serious questions.

The first one I would ask is why y axis does not start at 0 - this makes fluctuations look bigger than they should be, typical trick.

Secondly there was no need to make secondary axis because scale there is not that far off the first one - very close in fact, and thus it is also another trick for which people who make such charts should be fired on the spot: Gordon Brown will hire them to do his statistical presentations.

Milan is right - normally secondary axis is used when units on it are totally different in principle, say first chart is measured in dollars where as second is measured in tons of products, it is also useful when difference between minmax values of charts is huge or when you want exponential axis, ie: one axis could be power of earthquake using normal rank, but the other is actual power released - you need exponential axis for that since ranking is already build in exponenta.

[milanbenes]

no it was a serious question,

the chart has two x axis, one on the right and one on the left and both are scaled differently which makes the lines valueless as the trends are not making a comparison against the same scale

Milan.