Tag Archives: statistics

Recognising phoney on-line reviews

Sur­prise, sur­prise! The mar­ket­ing val­ue of pos­i­tive on-line user-reviews has cre­at­ed an indus­try of liars-for-hire, ready to pimp any prod­uct with phoney “user” endorse­ments. Hotels, cafs, pub­lish­ers, music labels and (hor­ror!) even blog­gers can read­i­ly find cut-price on-line tes­ti­mo­ni­al­ists to pimp their prod­uct on sites like fiverr.com. Unfor­tu­nate­ly, there’s a good chance we’ll fail to […]

Temperature trends in Garnaut 2011

I remain uncon­vinced by Dr Garnaut’s argu­ments that glob­al aver­age tem­per­a­ture trends in the past decade should, if any­thing, add to a sense of alarm about man-made “green­­house-gas” emis­sions. There is much in the updat­ed review that I have not had an oppor­tu­ni­ty yet to con­sid­er. (Update: There’s a more com­pre­hen­sive cri­tique of the updat­ed […]

Trade ‘imbalances’ are misleading

Alexan­dro Jara, the Deputy Direc­­tor-Gen­er­al of WTO “[R]elying on con­ven­tion­al trade sta­tis­tics gives a dis­tort­ed pic­ture of trade imbal­ances between coun­tries. As we saw when look­ing at the Chi­nese con­tent of the iPad, what counts is not the imbal­ances as mea­sured by gross val­ues of exports and imports, but how much val­ued added is embed­ded […]

Offensive teenagers

Offenders By Age And Sex

Accord­ing to the Aus­tralian Bureau of Sta­tis­tics more than 1 in 12 teenage males com­mits an offense that comes to police notice.

Rahmstorf rebuffed

The Pots­dam Insti­tute physi­cist whose 2007 paper Ross Gar­naut relied on for his asser­tion that “on the bal­ance of prob­a­bil­i­ties” CO2-dri­ven warm­ing was accel­er­at­ing dan­ger­ous­ly, has been exposed as a sci­en­tif­ic gad­fly.

At the time of the pub­li­ca­tion of Garnaut’s inter­im report, sev­er­al well-qual­i­fied scep­tics dis­put­ed Rahmstorf’s pro­je­tions, includ­ing David Stock­well, Lucia Lil­je­gren and Steve McIn­tyre with strong sup­port from for­mer Aus­tralian sta­tis­ti­cian Ian Cas­tles. Ian also kind­ly sup­port­ed my request to the Sta­tis­ti­cal Soci­ety of Aus­tralia to eval­u­ate the Rahm­storf method­ol­o­gy in the inter­ests of bet­ter informed pub­lic debate on Garnaut’s rec­om­men­da­tions (they even­tu­al­ly declined).

Now, A UK Met Office researcher and oceanog­ra­phers have harsh­ly crit­i­cised Ste­fan Rahm­storf for his extrav­a­gant pre­dic­tion that warm­ing would lead to sea-lev­el ris­es of 1.88 meters by the end of the cen­tu­ry.

Crit­ic Simon Hol­gate, a sea-lev­el expert at the Proud­man Oceano­graph­ic Lab­o­ra­to­ry, Mersey­side, has writ­ten to Sci­ence mag­a­zine, attack­ing Pro­fes­sor Rahmstorf’s work as ‘sim­plis­tic’.

‘Rahmstorf’s real skill seems to be in pub­lish­ing extreme papers just before big con­fer­ences like Copen­hagen, when they are guar­an­teed atten­tion,’ Dr Hol­gate said.” Extract from Sea-lev­el the­o­ry cuts no ice | The Aus­tralian

Inter­est­ing to note Rahmstorf’s weasly response to the crit­i­cism, report­ed at the end of The Australian’s sto­ry.

The ABS explains “trend estimates”

I’ve argued sev­er­al times that the broad­cast media has mis­rep­re­sent­ed quar­ter­ly unem­ploy­ment by focus­ing on (the scari­er of either) the raw or sea­son­al­ly-adjust­ed data. Although there’s a great deal to be said for just eyballing a trend when you have a long time-series that accu­rate­ly sam­ples an ‘atom­ic’ phe­nom­e­non, trends in a series with sta­tis­ti­cal­ly sim­ple (‘nor­mal­ly dis­trib­uted’) vari­abil­i­ty, like unem­ploy­ment, can rea­son­ably be rep­re­sent­ed by a smoothed trend esti­mate.

Trend esti­mates are obtained by smooth­ing the sea­son­al­ly adjust­ed esti­mates, with an assump­tion that the irreg­u­lar com­po­nent is ran­dom and nor­mal­ly dis­trib­uted. Dis­tor­tion of the trend esti­mate will occur in the pres­ence of an unusu­al event, and if no cor­rec­tion for the impact is intro­duced, then the trend esti­mate can be mis­lead­ing.” Extract from 1350.0 — Aus­tralian Eco­nom­ic Indi­ca­tors, Aug 2009

As the ABS release argues, this pro­ce­dure goes wrong when it’s no longer ‘busi­ness as usu­al.’ But the assump­tion that the world today is just like the world yes­ter­day is both pru­dent and pret­ty suc­cess­ful, espe­cial­ly at the lead­ing edge of a trend with estab­lished vari­a­tion. Cor­re­la­tion of smoothed data series is a dif­fer­ent mat­ter: it may be a species of fraud.

U.S. and Global Temperatures: a correction

Corrected GISS record shows 1934 as the hottest year

Dr Gavin Schmidt, a cli­mate mod­el­er at NASA and a prin­ci­pal author of the Real Cli­mate weblog, has emailed me to point out an error (mine) in my review of Ian Plimer’s Heav­en + Earth.

I said that I had learned from Ian Plimer that NASA had reversed it’s claim that the ten years fol­low­ing 1995 were the hottest ten years of the cen­tu­ry when Steven McIn­tyre showed that the record belonged to 1934. But, as Dr Schmidt points out, 1934 was the hottest year in the GISS records only for the Unit­ed States, not in the glob­al GISS records. He goes on to accuse Prof. Plimer of ‘lying’ to his read­ers about this, and oth­er, mat­ters.

I am grate­ful to Dr Schmidt for point­ing out this error, which I have cor­rect­ed. But I still share Ian Plimer’s amply-doc­u­ment­ed con­clu­sion that there is no rea­son for alarm about the slight warm­ing that undoubt­ed­ly took place over a fifty year peri­od from the 1940s.