Modelling Heterogeneous Dispersion in Marginal Models for Longitudinal Proportional Data

Size: px
Start display at page:

Download "Modelling Heterogeneous Dispersion in Marginal Models for Longitudinal Proportional Data"

Transcription

1 Bmetrcal Jurnal 46 (24) 5, DOI:.2/bmj.252 Mdellng Hetergeneus Dspersn n Margnal Mdels fr Lngtudnal Prprtnal Data Peter X.-K. Sng*;, Zhengu Qu 2, and Mng Tan 3 Department f Mathematcs and Statstcs, Yrk Unversty, Trnt, ON, Canada M3J P3 2 B.C. Research Insttute fr Chldren s and Wmen s Health, Vancuver, BC, Canada V5Z 4H4 3 Greenebaum Cancer Center N9E7, Unversty f Maryland, Baltmre, MD 22, USA Receved 24 July 2, revsed 22 Octber 23, accepted 5 June 24 Summary Cntnuus prprtnal data s cmmn n bmedcal research, e.g., the pre-pst therapy percent change n certan physlgcal and mlecular varables such as glmerular fltratn rate, certan gene expressn level, r telmere length. As shwn n (Sng and Tan, 2) such data requres methds beynd the cmmn generalsed lnear mdels. Hwever, the rgnal margnal smplex mdel f (Sng and Tan, 2) fr such lngtudnal cntnuus prprtnal data assumes a cnstant dspersn parameter. Ths assumptn f dspersn hmgenety s mpsed manly fr mathematcal cnvenence and may be vlated n sme stuatns. Fr example, the dspersn may vary n terms f drug treatment chrts r fllw-up tmes. Ths paper extends ther rgnal mdel s that the hetergenety f the dspersn parameter can be assessed and accunted fr n rder t cnduct a prper statstcal nference fr the mdel parameters. A smulatn study s gven t demnstrate that statstcal nference can be serusly affected by mstakenly assumng a varyng dspersn parameter t be cnstant n the applcatn f the avalable GEEs methd. In addtn, resdual analyss s develped fr checkng varus assumptns made n the mdellng prcess, e.g., assumptns n errr dstrbutn. The methds are llustrated wth the same eye surgery data n (Sng and Tan, 2) fr ease f cmparsn. Key wrds: Cntnuus prprtns; Generalsed lnear mdels; GEEs; Lngtudnal data; Resdual analyss; Smplex dstrbutn; Varyng dspersn. Intrductn The cncept f dspersn parameter s a famlar ne n generalsed lnear mdels (GLMs). The dspersn parameter f a nrmal dstrbutn s smply ts varance; and the dspersn parameter f Pssn dstrbutns s always equal t, whch s the rat f varance t mean and where the verdspersn ccurs when such a rat s larger than. Dspersn mdels (Jørgensen, 997), as an extensn f the GLMs, nclude dspersn parameters descrbng the dstrbutnal shape, whch s beynd what the lcatn r mean parameter alne can descrbe. The smplex dstrbutn f Barndrff-Nelsen and Jørgensen (99) fr the errr term represents a specal dspersn mdel, and s useful fr mdellng cntnuus prprtnal data. Based n ths dstrbutn, Sng and Tan (2) develped a margnal mdel fr lngtudnal cntnuus prprtnal data, whch was used t analyse an eye surgery data. Smlar t Lang and Zeger s margnal mdels e.g. Dggle et al. (22), Sng and Tan (2) assumed a cnstant dspersn n ther mdel and ther fcus s n mdellng the trend cmpnent. A techncal advantage by settng a cnstant dspersn parameter s that, as shwn n Lang and Zeger s GEE (986) apprach, regressn ceffcents can be separately estmated frm the dspersn parameter. Ths s because the GEE can factrse a cnstant dspersn ut the estmatng equatn. * Crrespndng authr: e-mal: sng@mathstat.yrku.ca # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

2 Bmetrcal Jurnal 46 (24) 5 54 Hwever, n practce the assumptn f hmgeneus dspersn may be questnable. Fr example, the magntude f dspersn may vary acrss drug treatment chrts due t dfferent rates f dsease prgressn r ver dfferent fllw-up tmes due t dfferent envrnmental expsures. It s clear that the margnal pattern f a ppulatn depends nt nly n ts averaged trend but als n ts dspersn characterstcs, as descrbed by the dspersn mdels. Therefre, ncrpratng varyng dspersn n the mdellng prcess allws us t assess the hetergenety f dspersn and t develp a smultaneus nference fr the entre margnal mdels cncernng bth trend and dspersn cmpnents. Such an access t the prfle f the dspersn parameter s mprtant, as shwn n ur smulatn studes n Sectn 4, mstakenly assumng a varyng dspersn t be cnstant n the applcatn f GEE methd culd cause sme serus prblems n statstcal nference. Fr example, the asympttc nrmalty thery fr the estmatrs may n lnger be vald, and ths thery s crucal t test fr statstcal sgnfcance fr the effects f sme cvarates f nterest. In addtn, a prper estmatn fr the dspersn parameter s appealng, fr example, n resdual analyss, where a standardsatn fr resduals s usually taken t stablse ther varances. The cmputatn f standardsed resduals always asks fr an apprprate estmate f the dspersn parameter. In ths paper, we prpse a new margnal mdel that cnssts f three cmpnents t be mdeled: the ppulatn-averaged effects, the dspersn pattern, and the crrelatn. In the cntext f lngtudnal data analyss, the frst versn f generalsed estmatng equatn apprach, knwn as f GEE, s prpsed by Lang and Zeger (986) and later extended by Prentce and Zha (99) t nclude a set f estmatng equatns n crrelatn parameters, referred t GEE2 n the lterature. As a matter f fact, estmatng a varyng dspersn parameter can be easly ncrprated wth the GEE2 usng the mean-varance relatnshp f the classcal GLMs r the expnental dspersn famly dstrbutns. Hwever, the mean-varance relatnshp s n lnger vald fr the smplex dstrbutn, because t s nt an expnental dspersn famly dstrbutn. Therefre, n ths paper, we suggest t add anther set f estmatng equatns t deal wth the dspersn cmpnent thrugh a certan mment prperty dfferent frm the mean-varance relatnshp. The resultng estmatng equatns extend the currently ppular GEE2, althugh t s stll called as GEE2 n the present paper. Mdellng dspersn parameter has been cnsdered by many authrs fr dfferent mdels n the lterature. Amng thers, Smyth (989) dscussed generalsed lnear mdels wth varyng dspersn fr crss-sectnal data, and Artes and Jørgensen (2) prpsed a mdel fr the ndex parameter f dspersn mdels, attemptng t attack ths prblem wth an underlyng applcatn clsely related t vn Mses dstrbutn fr lngtudnal crcular data. We fund ther methd dd nt wrk well fr the smplex dstrbutn. Pak (992) prpsed an estmatn prcedure that extends Lang and Zeger s GEEs by allwng bservatns frm dstrbutns wth dfferent dspersn parameters. Hwever, Pak s prcedure s nt applcable fr the smplex dstrbutn, because there s n clsed frm expressn fr the varance f the dstrbutn. By utlsng a certan mment prperty f the smplex dstrbutn, we cme up wth a dfferent slutn frm thse gven by Artes and Jørgensen (2) and Pak (992). In fact, because f dfferent perspectves and mdels, ur estmatng equatn fr the dspersn parameter f the smplex dstrbutn s smpler and numercally mre effcent than thers. The rest f the paper s rgansed as fllws. Sectn 2 presents dspersn margnal mdels wth varyng dspersn. An extended GEE2 s presented n Sectn 3, and Sectn 4 gves a smulatn study that demnstrates the mprtance f mdellng the dspersn parameter t cnduct a prper statstcal analyss n the presence f hetergeneus dspersn. Sectn 5 dscusses mdel dagnstcs thrugh resdual analyss. The prpsed methds are appled t re-analyse the eye surgery data n Sectn 6. Fnally we cnclude wth sme remarks. 2 Margnal Mdels T develp margnal smplex mdels fr lngtudnal cntnuus prprtnal data wth varyng dspersn, frst let y j, j ¼ ;...; n be the sequence f bserved repeated measurements n the th f m subjects, and t j, j ¼ ;...; n, be the sequence f crrespndng tmes n whch the measurements are taken n each subject. Asscated wth each y j are the values, x jk, k ¼ ;...; p, fp cvarates r expla- # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

3 542 P. X.-K. Sng et al.: Mdellng dspersn natry varables. We assume that y j are realsatns f randm varables Y j whch fllw smplex dstrbutns Y j S ðm j ; s 2 j Þ, where m j 2ð; Þ are the mean parameters and s2 j > are the dspersn parameters, and bth may be specfed as functns f cvarates. The densty functn f the smplex dstrbutn s, suppressng ndces, gven by h =2 pðy; m; s 2 Þ¼ 2ps 2 fyð yþg 3 exp dðy; mþ ; y 2ð; Þ ; 2s2 where d s the unt devance, ðy mþ 2 dðy; mþ ¼ yð yþ m 2 ð mþ 2 ; and ts unt varance functn s vðmþ ¼m 3 ð mþ 3. See (Jørgensen, 997) fr mre detals. Let Y ¼ðY ;...; Y n Þ > ; x j ¼ð; x j ;...; x jp Þ > : We assume that Y ;...; Y m are ndependent. A margnal smplex mdel cnssts f three cmpnents gven as fllws. The frst cmpnent s a mdel t descrbe the ppulatn-averaged effects, where the mean parameter m j depends n the tmevaryng cvarates x j va a generalsed lnear mdel f the frm hðm j Þ¼x > j b ðþ where h s a knwn lnk functn and b ¼ðb ;...; b p Þ > s the regressn ceffcents t be estmated. The lnk functn s usually chsen t be the lgt lnk functn that maps the untary nterval t ð ; Þ. The secnd cmpnent s a mdel t descrbe the pattern f dspersn parameter s 2 j as a functn f cvarates z j (maybe a subset f x j ), gven by gðs 2 j Þ¼z> j g ð2þ where g s a knwn lnk functn and g ¼ðg ;...; g r Þ > wth g crrespndng t the ntercept term. T express the dspersn as f a multplcatve frm, the lgarthm lnk functn s used t btan a lg-lnear mdel and hence s 2 j ¼ exp Qr ðz> j gþ¼ ðe g k Þ z jk ¼ e g Q r ðe g k Þ z jk : k¼ k¼ The thrd cmpnent s fr mdellng crrelatn structure. The crrelatn between Y j and Y k s a functn f the lcatn parameters and perhaps f addtnal parameters, a ¼ða ;...; a q Þ >, namely, crr ðy j ; Y k Þ¼qðm j ; m k ; aþ ð3þ where qðþ s a knwn functn. Varus types f crrelatn structures may be used fr the q functn. Amngst thers, three cmmnly used n the analyss f lngtudnal data are the exchangeable, AR() and m-dependence crrelatns. It s nted that the justfcatn fr a chce f a crrelatn structure s n general a dffcult task due t lttle nfrmatn ver tme avalable. Hwever the Lang and Zeger s GEE apprach fr cnsstent parameter estmatn enjys the rbustness aganst msspecfcatn f crrelatn structure and hence has yelded ppularty n lngtudnal data analyss. 3 GEEs fr Parameter Estmatn Dente the mean vectr f subject by m ¼ðm ;...; m n Þ >. Let the scre vectr fr subject be u ¼ðu ;...; u n Þ > ; wth u j ¼ 2 d ðy j ; m j Þ; # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

4 Bmetrcal Jurnal 46 (24) and under the regularty cndtns, Eðu j Þ¼ and therefre Eðu Þ¼. Frm Sng and Tan (2), the varance f u j s gven by var ðu j Þ¼ s2 j 2 Efd ðy j ; m j Þg ¼ 3s 4 j s2 j m j ð m j Þ þ vðm j Þ : Fllwng Sng and Tan (2), let w ¼ dag fvðm j Þg u be the wrkng vectr, and let RðaÞ be an n n wrkng crrelatn matrx wth a q vectr f crrelatn parameters a. S wrkng cvarance matrx fr w s V ¼ dag =2 var ðw j Þ RðaÞdag =2 var ðw j Þ : Therefre Lang and Zeger s GEE fr the smplex margn crrespnds t the estmatng equatn fr b gven by w ðb; g; aþ ¼ Pm D > A V w ¼ ; ð4þ where A ¼ dag fs 2 j vðm j Þ var ðu j Þg and D > > =@b. Fllwng Prentce and Zha (99), the GEE2 s frmed by addng an addtnal set f estmatng equatns fr the crrelatn parameters based n the standardsed scre resduals, defned by u j r j ¼ pffffffffffffffffffffffffffffff ¼ var ðu j Þ s j u j q ffffffffffffffffffffffffffffffffffffffffffffffffff : 2 Ed ðy j ; m j Þ It s easy t see that such scre resduals satsfy mment prpertes f Eðr j Þ¼, var ðr j Þ¼ and Eðr j r j Þ¼crr ðu j ; u j Þ¼crr ðw j ; w j Þ: The estmatng equatn fr the crrelatn parameter a then takes the frm w 3 ðb; g; aþ ¼ > H ðr Þ¼; ð5þ where r ¼ðr r 2 ; r r 3 ;...; r n r n Þ > ; H s a wrkng cvarance matrx and h ¼ Eðr Þ. The extended GEE2 cnssts f the equatns (4), (5), and an estmatng equatn fr the dspersn cmpnent gven as fllws, w 2 ðb; g; aþ ¼ S ðd s Þ¼ ; ð6þ where d ¼ðdðy ; m Þ;...; dðy n ; m n ÞÞ >, S s a wrkng cvarance matrx, and s ¼ Eðd Þ ¼ðs 2 ;...; s2 n Þ >. Nte that here we use the squared devance resduals, rather than the squared Pearsn resduals gven n Pak (992), t frm the thrd sets f estmatng equatns. The Crwder ptmal matrx fr S (Crwder, 987) s n fact the cv ðd Þ whch s n general nt easy t btan. A smple chce f S s the dentty matrx, leadng t the methd f mments estmatr fr g. Perhaps a better chce fr S s a dagnal matrx wth dagnal elements equal t the varances var fdðy j ; m j Þg ¼ 2ðs 2 j Þ2. See the appendx fr the prf f ths frmula n detal. Ths ndcates a gamma type f mean-varance relatn, that s, the unt varance functn s equal t the squared mean. Wth ths chce, the estmatng equatn wll effectvely prduce the quas-lkelhd estmatr f g as des the gamma regressn (Wedderburn, 974). Let q ¼ðb; g; aþ be the vectr f parameters t be estmated va the extended GEE2 fr whch the estmates are btaned by smultaneusly slvng the jnt equatns, 2 3 w ðb; g; aþ UðqÞ ¼Uðb; g; aþ ¼4 w 2 ðb; g; aþ 5 ¼ : (7) w 3 ðb; g; aþ # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

5 544 P. X.-K. Sng et al.: Mdellng dspersn It s clear that the estmatng equatn UðqÞ ¼ s unbased, namely EUðqÞ ¼. Hence t fllws frm the standard thery f estmatng equatns that under sme mld regularty cndtns, the estmatr ^q ¼ð^b; ^g; ^aþ s cnsstent and m =2 ð^q qþ s asympttcally multvarate Gaussan wth zer mean and cvarance matrx f the frm lm mj ðqþ, where JðqÞ s the Gdambe nfrmatn matrx m gven by JðqÞ ¼S > R S: Detals f the senstvty matrx S ¼ Ef@UðqÞ=@q > g and f the varablty matrx R ¼ EfUðqÞU > ðqþg are gven n the appendx. Usng the Newtn-scrng algrthm, the slutn ^q fr the jnt equatn (7) can be btaned numercally by teratvely updatng the q values as fllws, q ðkþþ ¼ q ðkþ S U q ðkþ : 4 Smulatn Study T demnstrate the mprtance f prperly analysng the lngtudnal data n the presence f hetergeneus dspersn, we cnduct a smulatn study where the prprtnal data y S ðm ; s 2 Þ; ¼ ;...; 5, were generated ndependently accrdng t the fllwng margnal mdels: lgt ðm Þ¼ b þ b T þ b 2 S ; lg ðs 2 Þ¼g þ g T ; where cvarates T and S are varables f treatment grups ndcated by (,, ), and llness severty scre ranged n (,, 2, 3, 4, 5, 6) that s randmly assumed t each subject by a bnmal dstrbutn B (6,.5). Fr smplcty, we manly nvestgated hw the parameters b j s representng the ppulatnaveraged effects wuld be affected by the stuatn f the dspersn parameter. S, we cnsdered nly the ndependence crrelatn structure, fr whch we were able t smulate data. We tk three equally szed treatment grups, each wth 5 subjects. Usng the ntatn abve, we yeld x ¼ð; T ; S 3Þ > and z ¼ð; T Þ > n whch the severty cvarate was centralsed by the md-scre 3. Mrever, the true values were assgned as ðb ; b ; b 2 Þ¼ð:5; :5; :5Þ, ðg ; g Þ¼ð3; 2Þ. We ran the regressn ver 2 replcatns, and the crrespndng summares are lsted belw. Table reprts the summary statstcs f the parameter estmates frm the extended GEE2 apprach prpsed n the paper wth hetergeneus dspersn. These statstcs nclude mean pnt estmate, 2.5th and 97.5th percentles, emprcal standard devatn and mean standard errr fr each f the fve parameters. When the mdel wth the hmgeneus dspersn was used t ft the smulated data, the mean estmate f lg ðs 2 Þ was 9.2 wth the emprcal standard devatn equal t 8.6, cnsderably larger than the average standard errr.2 btaned frm the sandwch asympttc varance estmatr. Table Summary statstcs f the estmates, based n 2 replcatns generated frm the hetergenety mdel. True Value Hetergeneus Dspersn Hmgeneus Dspersn Mean (2.5%, 97.5%) Stdev* Stderr y Mean (2.5%, 97.5%) Stdev Stderr b (.5).4986 (.367,.668) (.275,.588) b (.5).546 (.6697,.3493) (.947,.9598) b 2 (.5).53 (.3973,.659) (.7345,.79) g ( 3.) ( , ) g ( 2.) 2.88 (.78, 2.284) * Emprcal standard devatn; y Mean standard errr # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

6 Bmetrcal Jurnal 46 (24) Frm Table, we learned: () The pnt estmates ^b ; ¼ ; ; 2 n the pupulatn-averaged effects mdel () frm bth appraches are relatvely clse t each ther, althugh the mdel wth the hmgeneus dspersn prduces a lttle larger devatn frm the true values than the mdel wth the hetergeneus dspersn. () The 95% emprcal cnfdence ntervals frm the tw mdels have substantally dfferent cverage, zer beng ncluded n the ntervals gven by the hmgenety mdel as ppsed t zer beng excluded n thse gven by the hetergenety mdel, fr all three b parameters. Ths suggests that the hmgenety mdel lses ts pwer f dentfyng sme mprtant cvarates n the presence f hetergeneus dspersn. () The values f the emprcal standard devatn and the standard errr are very smlar n the hetergenety mdel, but clearly dfferent n the hmgenety mdel. Ths ndcates that the asympttc nrmalty thery fr the estmatrs frm the hmgenety mdel may be n lnger vald. T vsualse ths, we pltted the estmated denstes ver the 2 estmates fr each parameter n Fgure. Fgure Estmated denstes f the mdel parameters ver 2 replcatns usng data generated frm the hetergenety mdel. # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

7 546 P. X.-K. Sng et al.: Mdellng dspersn Fgure ndcates that fr each parameter, the estmates frm the hetergenety mdel are evenly dstrbuted alng the parameter space and clearly frm a bell-shaped densty. In cntrast, the estmates frm the hmgenety mdel ccurs mre frequently n tal areas and clearly frm a heavy-taled densty. The densty fr the estmatr f g frm the hmgenety mdel has an extremely lng tal n the rght. In cnclusn, the asympttc nrmalty fr the estmatrs frm the hmgenety mdel s serusly n questn. Cnversely, we cnducted anther smulatn n that the true mdel had the hmgeneus dspersn. In partcular, data were generated smlarly as n the frst smulatn, except that nw the dspersn mdel s cnstant lg ðs 2 Þ¼g. The true values f b parameters are the same as abve, and set g ¼ 4, whch leads t a large dspersn arund 55. Table 2 gves the summary statstcs ver 2 replcatns. Table 2 Summary statstcs f the estmates, based n 2 replcatns generated frm the hmgenety mdel. True Value Hetergeneus Dspersn Hmgeneus Dspersn Mean (2.5%, 97.5%) Stdev* Stderr y Mean (2.5%, 97.5%) Stdev Stderr b (.5).57 (.2938,.788) (.2942,.75) b (.5).578 (.6943,.2885) (.6992,.294) b 2 (.5).59 (.3623,.688) (.369,.6857) g ( 4.) ( 3.726, 4.27) ( , 4.249) g (.).99 (.2874,.2889) Evdentally, Table 2 ndcates that the estmates frm the tw mdels are very clse, the null hypthess H : g ¼ cannt be rejected at the sgnfcance level :5 under the hetergenety mdel. As expected, the estmated denstes (nt shwn n the paper) f the parameters are very smlar between the tw mdels, and they are all very alke t nrmal densty curves. In summary, when a cnstant dspersn assumptn s n dubt, the hetergenety mdel seems t be necessary and advantageus t make prper statstcal nference. 5 Resdual Analyss * Emprcal standard devatn; y Mean standard errr We prpse t use tw types f resduals t frm dagnstcs fr the key mdel assumptns: () margnal dstrbutns, () lnk functns, and () the wrkng crrelatn structure. The frst ne s the standardsed scre resduals r j gven n (5), and the ther s the regular standardsed Pearsn pffffffffffffffffffffffffffffffff resduals, e j ¼ðy j m j Þ= var ðy j Þ, where var ðy j Þ has n clsed frm expressn as t nvlves the ncmplete gamma functn. See Jørgensen (997) fr the detals. The sample cunterpart f r j r e j s btaned by replacng parameters by ther crrespndng estmates, dented by ^r j r ^e j, accrdngly. Lke mst resdual analyses, ur resdual analyss belw s useful t detect strng sgnals asscated wth certan mdel assumptn vlatn. The smplex dstrbutn assumptn can be checked by the plt f ^e j aganst ^m j, whch ams t examne the mean-varance relatn. If ths assumptn s true, then var ðe j Þ¼, ndependent f mean m j. Therefre, pnts n the plt shuld randmly scatter arund the hrzntal lne at zer (the expectatn f resduals), wth apprxmately 95% pnts n the hrzntal band between 2 and 2. Any apparent departure frm ths wuld suggest ether a vlatn f the assumptn n dstrbutn r prbably a pr mdel ft. A seres f further nvestgatns are needed t dentfy whch factr s respnsble fr such departure. Ths apprach wuld becme mre relable as s 2 becmes large, because the mean-varance relatn becmes dmnated by mð mþ, a case smlar t that f a bnmal dstrbutn. # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

8 Bmetrcal Jurnal 46 (24) Fllwng McCullagh and Nelder s (989), we use the plt f the adjusted dependent varable s j aganst the lnear predctr ^h j t check the chsen lnk functn. In ur settng, defne ( ) 3s 4 =2 j s j ¼ hðm j Þþ m j ð m j Þ þ s2 j uðy j; m vðm j Þ Þ; j ¼ ;...; n j ; ¼ ;...; m : Clearly, Eðs j Þ¼hðm j Þ snce Eðu j Þ¼, and var ðs j Þ¼Efs j hðm j Þg 2 ¼. If the lnk functn s apprprate, the plt f the estmates ^s j aganst ^h j ¼ x > ^b j shuld shw a straght lne wth apprxmately 95% pnts fallng nt a band wth the upper and lwer lmts f ^h j 2. As n generalsed lnear mdels, ths plt des nt suggest the best lnk functn fr the mdel, but rather nly gves an nfrmal check fr any strng vlatn f the used lnk. Althugh t s dffcult t mdel the true crrelatn structure f lngtudnal data, apprxmately crrect crrelatn structures allw regressn ceffcents t be estmated mre effcently. Thus, t s mprtant t assess the apprprateness f the wrkng crrelatn used n GEEs va resdual analyss. Nte that crr ðr j ; r j Þ¼crr ðw j ; w j Þ; mplyng that the true crrelatn f varable w j s equal t that f the standardsed scre resduals r j. Sme explratry prcedures presented n Sectn 3.4 f Dggle et al. (22) can be adpted fr w j s t examne the crrelatn f data. 6 An Example In ths sectn we re-analyse the phthalmlgcal data n the use f ntracular gas n retnal repar surgeres (Meyers et al., 992), wth a specal fcus n the hetergeneus dspersn. A prmary analyss f the data assumng the hmgeneus dspersn was dne by Sng and Tan (2). Brefly, the study was t nvestgate the decay curse f the ntracular gas n retnal repar surgeres prspectvely n 3 patents. The gas was njected nt the eye befre surgery and patents were fllwed three t eght (average f 5) tmes ver a three-mnth perd. The respnse varable y j was the percent f gas left n the eye recrded as prprtn (a percent). The questn was f the dsappearance f the gas s related t ther cvarates such as the cncentratn f the gas used. Sng and Tan (2) mdelled the gas vlume drectly usng a margnal mdel. Wth ur prpsed methd, we are able t test f the hmgeneus dspersn s true, and f nt s the mdel allws us t dentfy whch cvarates lead t hetergenety. T begn, the ppulatn-averaged effects mdel n Sng and Tan (2) s lgt ðm j Þ¼b þ b lg ðt j Þþb 2 lg 2 ðt j Þþb 3 x j ð8þ where t j s the tme cvarate f days after the gas njectn, and x j s the cvarate f gas cncentratn levels equal t, and, crrespndng t the cncentratn levels f 5%, 2% and 25%, respectvely. T ths mdel, the cmpnents f the estmatng functn w specfed by (4) are gven as fllws. D > ¼ X > dag fm j ð m j Þg ; D > A ¼ X > dag f3s 2 j vðm j Þþm j ð m j Þg; where X s f n 3 dmensn and ts jth rw s ð; lg ðt j Þ; lg 2 ðt j Þ; x j Þ, and var ðw j Þ¼s 2 j vðm j Þf3s2 j m2 j ð m j Þ2 þ g: Clearly the crrespndng senstvty matrx s S ¼ Pm D > A V A D. Als as ndcated n ther paper, AR() dependence seemed t ft the data the best, s ur analyss nly cncerns ths type f dependence, specfed as f the frst-rder ECM mdel, crr ðw j ; w j Þ ¼ exp ðajt j t j jþ, fr a <. When H s chsen t be the dentty matrx, the functn w 3 becmes FðaÞ ¼ Pm c > ðr h Þ ¼ # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

9 548 P. X.-K. Sng et al.: Mdellng dspersn Table 3 Estmates, standard errrs and rbust z-statstcs frm the hetergeneus dspersn mdel fr the eye surgery data. Parameter b b b 2 b 3 g g g 2 a Estmate Stderr? z-statstc *Standard Errr where c ¼½jt t 2 j exp ð ajt t 2 jþ;...; jt n t n j exp ð ajt n t n jþš > and the crrespndng senstvty matrx S 33 ¼ Pm c > c. The mdel that addresses the hetergenety n tw cvarates f tme and gas cncentratn level takes the fllwng frm lg ¼ g þ g lg ðt j Þþg 2 x j : ð9þ s 2 j We ran the Newtn-scrng algrthm gven n Sectn 3 and fund estmates and standard errrs that are lsted n Table % 2% 25% Lg-dspersn Days Fgure 2 Ftted curves fr the pattern f hetergeneus dspersn ver tme acrss three treatment levels. # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

10 Clearly, bth cvarates f tme and treatment are sgnfcant factrs attrbuted t the hetergeneus dspersn n mdel (9). Fgure 2 dsplays the ftted curves fr the pattern f dspersn prfle ver tme acrss three dfferent gas cncentratn levels. Based n the mdel wth the tme-varyng dspersn, ur fndngs fr ther parameters are very smlar t thse n Sng and Tan (2). Smlar t Sng and Tan (2), we fund that the quadratc tme term lg 2 ðt j Þ s sgnfcant, that the lnear tme lg ðt j Þ s nt sgnfcant, and that the gas cncentratn cvarate s margnally nsgnfcant, at the sgnfcance level.5. Als, The estmated lag- autcrrelatn ^q ¼ e^a ¼ :575ð:6Þ and ts z-statstc s , suggestng that q s sgnfcantly dfferent frm zer. In cntrast t the smulatn study, here we dd nt see dramatc dfferences between the results frm the hetergenety and hmgenety mdels. We gave the reasn as fllws. In the smulatn study we chse the ntercept and slpe parameters t be cmparable, respectvely 3 and 2, s that a change n the cvarate wuld greatly affect the sze f dspersn. Therefre, the results frm the hmgenety and hetergenety mdels were evdently dfferent. Hwever, n the data analyss the ntercept dmnates the cntrbutn t the dspersn ver the tw slpe ceffcents, mplyng that the verall dspersn remans mstly very large, and therefre n bg dfferences appeared n the results frm the tw types f mdels. Nw we cnsder the resdual analyss fr the abve mdel wth tme-varyng dspersn. Panel A f Fgure 3 shws the scatter-plt f the estmated standardsed Pearsn resduals ^e j s aganst the ftted mean values ^m j, t check the dstrbutn assumptn. The dashed lnes at 2 and 2 represent the asympttc 95% upper and lwer lmts, respectvely. The resduals seem t behave reasnably well as expected, nly three f them lyng utsde f the regn. The plt seems t be n agreement wth the smplex margnal dstrbutn. Panel B f Fgure 3 prvdes a rugh check f the lgt lnk functn used n the prpsed mdel, shwng the scatter-plt f the estmated adjusted dependent varables ^s j aganst the estmated lgt Bmetrcal Jurnal 46 (24) Pearsn resduals Adjusted dependent varable Ftted values Panel A: Checkng dstrbutn assumptn Lgt lnear predctr Panel B: Checkng lnk funktn Fgure 3 Dagnstc plts n the eye surgery data analyss. # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

11 55 P. X.-K. Sng et al.: Mdellng dspersn lnear predctr ^h j. The tw sld lnes stand fr the asympttc 95% cnfdent bands wthn whch almst 96% pnts are cntaned. Ths clearly supprts the lgt lnk functn assumptn. Checkng the wrkng crrelatn seems t be nntrval, snce the data are measured at rregular tme pnts and the resduals avalable at a gven tme are sparse. S we feel that the prpsed methd fr checkng the wrkng crrelatn may nt be relable here. Alternatvely, Dggle s vargram plt (Dggle, 99) may be used here t reach an apprprate cnclusn. Hwever ths s nt the fcus f the paper, and hence the detals are mtted. 7 Cncludng remarks In ths paper we develped an apprach t mdellng the hetergeneus dspersn parameter, relaxng the usual assumptn f cnstant dspersn n Lang and Zeger s margnal mdels. An extended versn f GEEs was prpsed t estmate the parameters n the mdel fr dspersn. Thrugh the analyss f the eye surgery data, we fund that the dspersn can be a functn ver fllw-up tme as well as treatment arm, and that the shape f margnal dstrbutns s tme-varyng n addtn t the tme-varyng lcatns. Ths prpsed methd mprves the mdellng f lngtudnal data and prvdes a tl fr better understandng the margnal prfles f the lngtudnal cntnuus prprtnal data. The extended GEEs n ths paper was develped under the assumptn f n mssng values n data. Snce mssng values ften ccur n lngtudnal studes n practce, t wuld be f great nterest t further extend the prpsed GEEs t cnduct data analyss wth mssng values. It s knwn that GEEs prduce cnsstent estmatrs fr the mdel parameters when mssng values are cmpletely randm and gnred n the analyss. Hwever, when data cntan randm mssng values r nfrmatve mssng values, the cnsstency fr the GEEs estmatrs s generally n lnger vald. Reslvng ths ssue has been an actve research tpc n the lngtudnal data analyss. Fr example, Rbns et al. (995) suggested the nverse prbablty weghted GEEs that prduce cnsstent estmates f the drp-ut prcess s prperly mdelled. Anther apprach suggested by Pak (997) s t mpute the mssng values by the cndtnal expectatn gven the bserved data. Mre references can be fund n Dggle et al. (22), Verbeke and Mlenberghs (2), r Zegler et al. (998). Acknwledgements The authrs are grateful t the tw referees fr ther valuable suggestns and cmments that led t an mprvement f the paper. The authrs thank Dr. Sanfrd Meyers fr makng hs data avalable fr nclusn as an example. The frst tw authrs research was supprted by the Natnal Scence and Engneerng Research Cuncl f Canada. Appendx A Gdambe nfrmatn matrx Ths sectn gves the cmpnents f Gdambe nfrmatn matrx needed fr cmputng the estmated standard errrs fr estmates and hence fr cnstructng Wald test statstcs. The senstvty matrx s a 3 3 blck matrx, S ¼ S S 2 S > S 2 S 22 S 23 A; S 3 S 32 S 33 where clearly S 2 ¼, S 3 ¼, and S 23 ¼. Als the blck S 2 ¼ because Eu j ¼. S n general the S matrx takes the frm S S S 22 A; S 3 S 32 S 33 # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

12 Bmetrcal Jurnal 46 (24) 5 55 and ts nverse matrx s S S S 22 A; S 33 S 3S S 33 S 32S 22 S 33 prvded that all dagnal blcks are nvertble. When the dstrbutn f r j r j s ndependent f the mean and dspersn parameters, bth S 3 and S 32 are. Therefre the matrx S becmes a blckdagnal matrx wth S ¼ Xm D > A V A D ; S 22 ¼ @g > and S 33 ¼ @a > : The varablty matrx R s als a 3 3 blck matrx, V V 2 V 3 V ¼ EfUðqÞU > ðqþg V 2 V 22 V 23 A: V 3 V 32 V 33 The nne blcks are detaled as fllws. V ¼ Efw w > g¼xm V 2 ¼ Efw w > 2 g¼xm V 3 ¼ Efw w > 3 g¼xm V 22 ¼ Efw 2 w > 2 g¼xm V 23 ¼ Efw 2 w > 3 g¼xm V 33 ¼ Efw 3 w > 3 g¼xm D > A V cv ðw ÞV A D ; D > A V cv ðw ; d > ; D > A V cv ðw ; r > > > S cv ðd Þ S S cv ðd ; r Þ H H cv ðr > : Because f symmetry, V 2 ¼ V > 2, V 3 ¼ V > 3, and V 32 ¼ V > 23. It s nted that cv ðw Þ¼dag fvðm j Þg cv ðu Þ dag fvðm j Þg, and an estmate f cv ðu Þ s btaned by pluggng the estmates ^m j and replacng cv ðu Þ by ^u ^u > n the expressn. The same apprach s appled t estmate the remanng blcks f V. # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

13 552 P. X.-K. Sng et al.: Mdellng dspersn B Prf f mean-varance relatn Ths sectn presents the prf fr the frmula var fdðy j ; m j Þg ¼ 2ðs 2 j Þ2, Y j S ðm j ; s 2 jþ. Frm the appendx f Sng and Tan (2), suppressng crdnates, EfdðY; mþg ¼ s 2, and hence t s suffcent t shw that Efd 2 ðy; mþg ¼ 3ðs 2 Þ 2. A smple algebra leads t Efd 2 ðy; mþg ¼ Z d 2 ðy; mþ pðy; m; s 2 Þ dy rffffffffff l ð þ xþ 4 ¼ 2p x 4 Z fx 3 2 þð 4xÞ x 2 þ 2xð3x 2Þ x 2 þ 2x 2 ð3 2xÞ x 3 2 þ x 3 ðx 4Þ x 5 2 þ x 4 x 7 2 g f ðx; x; lþ dx; where l ¼ =s 2 and ( ) f ðx; x; lþ ¼exp l ð þ xþ 2 ðx xþ 2 2 x 2 : x Usng frmulas (5.4) (5.43) f (Jørgensen, 997), we btan and Z Z x 3 2 f ðx; x; lþ dx ¼ x 7 2 f ðx; x; lþ dx ¼ 2p 2 l 2 x 3 ð þ xþ 4 þ 3lx 4 ð þ xþ 2 þ 3x 5 l l 2 ð þ xþ 5 2p 2 l 2 ð þ xþ 4 þ 3lxð þ xþ 2 þ 3x 2 l l 2 x 2 ð þ xþ 5 : Pluggng these results and thse frm Sng and Tan (2), we get E d 2 ðy; mþ ¼ 3 s 2 ð Þ 2. References Artes, R. and Jørgensen, B. (2). Lngtudnal data estmatng equatns fr dspersn mdels. Scandnavan Jurnal f Statstcs 27, Barndrff-Nelsen, O. E. and Jørgensen, B. (99). Sme parametrc mdels n the smplex. Jurnal f Multvarate Analyss 39, 6 6. Crwder, M. (987). On lnear and quadratc estmatng functns. Bmetrka 74, Dggle, P. J. (99). Tme Seres: A Bstatstcal Intrductn. Oxfrd Unversty Press, Oxfrd. Dggle, P. J. Heagerty, P., Lang, K.-Y. and Zeger, S. L. (22). The Analyss f Lngtudnal Data, 2nd ed. Oxfrd Unversty Press, Oxfrd. Jørgensen, B. (997). The Thery f Dspersn Mdels. Chapman Hall, Lndn. Lang, K.-Y. and Zeger, S. L. (986). Lngtudnal data analyss usng generalzed lnear mdels. Bmetrka 73, McCullagh, P. and Nelder, J. A. (989). Generalsed Lnear Mdels, 2nd ed. Chapman and Hall, Lndn. Meyers, S. M., Ambler, J. S., Tan, M., Werner, J. C. and Huang, S. S. (992). Varatn f perflurprpane dsappearance after vtrectmy. Retna 2, Pak, M. C. (992). Parametrc varance functn estmatn fr nnnrmal repeated measurement data. Bmetrcs 48, 9 3. Pak, M. C. (997). The generalzed estmatng equatn apprach when data are nt mssng cmpletely at randm. Jurnal f Amercan Statstcal Asscatn 92, Prentce, R. L. and Zha, L. P. (99). Estmatng equatns fr parameters n means and cvarances f multvarate dscrete and cntnuus respnses. Bmetrcs 47, # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

14 Bmetrcal Jurnal 46 (24) Rbns, J. M., Rtntzky, A. and Zha, L. P. (995). Analyss f semparametrc regressn mdels fr repeated utcmes n the presence f mssng data. Jurnal f the Amercan Statstcal Asscatn 4, Smyth, G. K. (989). Generalsed lnear mdels wth varyng dspersn. Jurnal f the Ryal Statstcal Scety, Seres B 5, Sng, P. X.-K. and Tan, M. (2). Margnal mdels fr lngtudnal cntnuus prprtnal data. Bmetrcs 56, Verbeke, G. and Mlenberghs, G. (2). Lnear Mxed Mdels fr Lngtudnal Data. Sprnger-Verlag, New Yrk. Wedderburn, R. W. M. (974). Quas-lkelhd functns, generalzed lnear mdels and the Gauss-Newtn methd. Bmetrka 6, Zegler, A., Kastner, C. and Blettner, M. (998). The generalsed estmatng equatns: an anntated bblgraphy. Bmetrcal Jurnal 4, # 24 WILEY-VCH Verlag GmbH & C. KGaA, Wenhem

ENGI 4421 Probability & Statistics

ENGI 4421 Probability & Statistics Lecture Ntes fr ENGI 441 Prbablty & Statstcs by Dr. G.H. Gerge Asscate Prfessr, Faculty f Engneerng and Appled Scence Seventh Edtn, reprnted 018 Sprng http://www.engr.mun.ca/~ggerge/441/ Table f Cntents

More information

Section 10 Regression with Stochastic Regressors

Section 10 Regression with Stochastic Regressors Sectn 10 Regressn wth Stchastc Regressrs Meanng f randm regressrs Untl nw, we have assumed (aganst all reasn) that the values f x have been cntrlled by the expermenter. Ecnmsts almst never actually cntrl

More information

Regression with Stochastic Regressors

Regression with Stochastic Regressors Sectn 9 Regressn wth Stchastc Regressrs Meanng f randm regressrs Untl nw, we have assumed (aganst all reasn) that the values f x have been cntrlled by the expermenter. Ecnmsts almst never actually cntrl

More information

Inference in Simple Regression

Inference in Simple Regression Sectn 3 Inference n Smple Regressn Havng derved the prbablty dstrbutn f the OLS ceffcents under assumptns SR SR5, we are nw n a pstn t make nferental statements abut the ppulatn parameters: hypthess tests

More information

A New Method for Solving Integer Linear. Programming Problems with Fuzzy Variables

A New Method for Solving Integer Linear. Programming Problems with Fuzzy Variables Appled Mathematcal Scences, Vl. 4, 00, n. 0, 997-004 A New Methd fr Slvng Integer Lnear Prgrammng Prblems wth Fuzzy Varables P. Pandan and M. Jayalakshm Department f Mathematcs, Schl f Advanced Scences,

More information

Chapter 7. Systems 7.1 INTRODUCTION 7.2 MATHEMATICAL MODELING OF LIQUID LEVEL SYSTEMS. Steady State Flow. A. Bazoune

Chapter 7. Systems 7.1 INTRODUCTION 7.2 MATHEMATICAL MODELING OF LIQUID LEVEL SYSTEMS. Steady State Flow. A. Bazoune Chapter 7 Flud Systems and Thermal Systems 7.1 INTODUCTION A. Bazune A flud system uses ne r mre fluds t acheve ts purpse. Dampers and shck absrbers are eamples f flud systems because they depend n the

More information

SIMULATION OF THREE PHASE THREE LEG TRANSFORMER BEHAVIOR UNDER DIFFERENT VOLTAGE SAG TYPES

SIMULATION OF THREE PHASE THREE LEG TRANSFORMER BEHAVIOR UNDER DIFFERENT VOLTAGE SAG TYPES SIMULATION OF THREE PHASE THREE LEG TRANSFORMER BEHAVIOR UNDER DIFFERENT VOLTAGE SAG TYPES Mhammadreza Dlatan Alreza Jallan Department f Electrcal Engneerng, Iran Unversty f scence & Technlgy (IUST) e-mal:

More information

Chapter 3, Solution 1C.

Chapter 3, Solution 1C. COSMOS: Cmplete Onlne Slutns Manual Organzatn System Chapter 3, Slutn C. (a If the lateral surfaces f the rd are nsulated, the heat transfer surface area f the cylndrcal rd s the bttm r the tp surface

More information

V. Electrostatics Lecture 27a: Diffuse charge at electrodes

V. Electrostatics Lecture 27a: Diffuse charge at electrodes V. Electrstatcs Lecture 27a: Dffuse charge at electrdes Ntes by MIT tudent We have talked abut the electrc duble structures and crrespndng mdels descrbng the n and ptental dstrbutn n the duble layer. Nw

More information

Shell Stiffness for Diffe ent Modes

Shell Stiffness for Diffe ent Modes Engneerng Mem N 28 February 0 979 SUGGESTONS FOR THE DEFORMABLE SUBREFLECTOR Sebastan vn Herner Observatns wth the present expermental versn (Engneerng Dv nternal Reprt 09 July 978) have shwn that a defrmable

More information

Fall 2010 Analysis of Experimental Measurements B. Eisenstein/rev. S. Errede. (n.b. for now, we do not require that k. vectors as a k 1 matrix: ( )

Fall 2010 Analysis of Experimental Measurements B. Eisenstein/rev. S. Errede. (n.b. for now, we do not require that k. vectors as a k 1 matrix: ( ) Fall 00 Analyss f Epermental Measrements B. Esensten/rev. S. Errede Let s nvestgate the effect f a change f varables n the real & symmetrc cvarance matr aa the varance matr aa the errr matr V [ ] ( )(

More information

Section 3: Detailed Solutions of Word Problems Unit 1: Solving Word Problems by Modeling with Formulas

Section 3: Detailed Solutions of Word Problems Unit 1: Solving Word Problems by Modeling with Formulas Sectn : Detaled Slutns f Wrd Prblems Unt : Slvng Wrd Prblems by Mdelng wth Frmulas Example : The factry nvce fr a mnvan shws that the dealer pad $,5 fr the vehcle. If the stcker prce f the van s $5,, hw

More information

Wp/Lmin. Wn/Lmin 2.5V

Wp/Lmin. Wn/Lmin 2.5V UNIVERITY OF CALIFORNIA Cllege f Engneerng Department f Electrcal Engneerng and Cmputer cences Andre Vladmrescu Hmewrk #7 EEC Due Frday, Aprl 8 th, pm @ 0 Cry Prblem #.5V Wp/Lmn 0.0V Wp/Lmn n ut Wn/Lmn.5V

More information

CHAPTER 3 ANALYSIS OF KY BOOST CONVERTER

CHAPTER 3 ANALYSIS OF KY BOOST CONVERTER 70 CHAPTER 3 ANALYSIS OF KY BOOST CONERTER 3.1 Intrductn The KY Bst Cnverter s a recent nventn made by K.I.Hwu et. al., (2007), (2009a), (2009b), (2009c), (2010) n the nn-slated DC DC cnverter segment,

More information

Lecture 12. Heat Exchangers. Heat Exchangers Chee 318 1

Lecture 12. Heat Exchangers. Heat Exchangers Chee 318 1 Lecture 2 Heat Exchangers Heat Exchangers Chee 38 Heat Exchangers A heat exchanger s used t exchange heat between tw fluds f dfferent temperatures whch are separated by a sld wall. Heat exchangers are

More information

Water vapour balance in a building moisture exposure for timber structures

Water vapour balance in a building moisture exposure for timber structures Jnt Wrkshp f COST Actns TU1 and E55 September 21-22 9, Ljubljana, Slvena Water vapur balance n a buldng msture expsure fr tmber structures Gerhard Fnk ETH Zurch, Swtzerland Jchen Köhler ETH Zurch, Swtzerland

More information

4DVAR, according to the name, is a four-dimensional variational method.

4DVAR, according to the name, is a four-dimensional variational method. 4D-Varatnal Data Assmlatn (4D-Var) 4DVAR, accrdng t the name, s a fur-dmensnal varatnal methd. 4D-Var s actually a smple generalzatn f 3D-Var fr bservatns that are dstrbuted n tme. he equatns are the same,

More information

element k Using FEM to Solve Truss Problems

element k Using FEM to Solve Truss Problems sng EM t Slve Truss Prblems A truss s an engneerng structure cmpsed straght members, a certan materal, that are tpcall pn-ned at ther ends. Such members are als called tw-rce members snce the can nl transmt

More information

Introduction to Electronic circuits.

Introduction to Electronic circuits. Intrductn t Electrnc crcuts. Passve and Actve crcut elements. Capactrs, esstrs and Inductrs n AC crcuts. Vltage and current dvders. Vltage and current surces. Amplfers, and ther transfer characterstc.

More information

Conduction Heat Transfer

Conduction Heat Transfer Cnductn Heat Transfer Practce prblems A steel ppe f cnductvty 5 W/m-K has nsde and utsde surface temperature f C and 6 C respectvely Fnd the heat flw rate per unt ppe length and flux per unt nsde and per

More information

Section 14 Limited Dependent Variables

Section 14 Limited Dependent Variables Sectn 14 Lmted Dependent Varables What s a lmted dependent varable? Our standard assumptn f an errr term that s nrmally dstrbuted cndtnal n the regressrs mples that the dependent varable can be (wth pstve

More information

Integrating Certified Lengths to Strengthen Metrology Network Uncertainty

Integrating Certified Lengths to Strengthen Metrology Network Uncertainty Integratng Certfed engths t Strengthen Metrlgy Netwrk Uncertanty Authrs: Jseph Calkns, PhD New Rver Knematcs je@knematcs.cm Sctt Sandwth New Rver Knematcs sctt@knematcs.cm Abstract Calbrated and traceable

More information

6. ELUTRIATION OF PARTICLES FROM FLUIDIZED BEDS

6. ELUTRIATION OF PARTICLES FROM FLUIDIZED BEDS 6. ELUTRIATION OF PARTICLES FROM FLUIDIZED BEDS Elutratn s the prcess n whch fne partcles are carred ut f a fludzed bed due t the flud flw rate passng thrugh the bed. Typcally, fne partcles are elutrated

More information

State-Space Model Based Generalized Predictive Control for Networked Control Systems

State-Space Model Based Generalized Predictive Control for Networked Control Systems Prceedngs f the 7th Wrld Cngress he Internatnal Federatn f Autmatc Cntrl State-Space Mdel Based Generalzed Predctve Cntrl fr Netwred Cntrl Systems Bn ang* Gu-Png Lu** We-Hua Gu*** and Ya-Ln Wang**** *Schl

More information

A Note on the Linear Programming Sensitivity. Analysis of Specification Constraints. in Blending Problems

A Note on the Linear Programming Sensitivity. Analysis of Specification Constraints. in Blending Problems Aled Mathematcal Scences, Vl. 2, 2008, n. 5, 241-248 A Nte n the Lnear Prgrammng Senstvty Analyss f Secfcatn Cnstrants n Blendng Prblems Umt Anc Callway Schl f Busness and Accuntancy Wae Frest Unversty,

More information

Spring 2002 Lecture #17

Spring 2002 Lecture #17 1443-51 Sprng 22 Lecture #17 r. Jaehn Yu 1. Cndtns fr Equlbrum 2. Center f Gravty 3. Elastc Prpertes f Slds Yung s dulus Shear dulus ulk dulus Tday s Hmewrk Assgnment s the Hmewrk #8!!! 2 nd term eam n

More information

Transient Conduction: Spatial Effects and the Role of Analytical Solutions

Transient Conduction: Spatial Effects and the Role of Analytical Solutions Transent Cnductn: Spatal Effects and the Rle f Analytcal Slutns Slutn t the Heat Equatn fr a Plane Wall wth Symmetrcal Cnvectn Cndtns If the lumped capactance apprxmatn can nt be made, cnsderatn must be

More information

PHYSICS 536 Experiment 12: Applications of the Golden Rules for Negative Feedback

PHYSICS 536 Experiment 12: Applications of the Golden Rules for Negative Feedback PHYSICS 536 Experment : Applcatns f the Glden Rules fr Negatve Feedback The purpse f ths experment s t llustrate the glden rules f negatve feedback fr a varety f crcuts. These cncepts permt yu t create

More information

Analytical Modeling of Natural Convection in Horizontal Annuli

Analytical Modeling of Natural Convection in Horizontal Annuli Analytcal Mdelng f Natural Cnvectn n Hrzntal Annul Peter Teertstra, M. Mchael Yvanvch, J. Rchard Culham Mcrelectrncs Heat Transfer Labratry Department f Mechancal Engneerng Unversty f Waterl Waterl, Ontar,

More information

Physics 107 HOMEWORK ASSIGNMENT #20

Physics 107 HOMEWORK ASSIGNMENT #20 Physcs 107 HOMEWORK ASSIGNMENT #0 Cutnell & Jhnsn, 7 th etn Chapter 6: Prblems 5, 7, 74, 104, 114 *5 Cncept Smulatn 6.4 prves the ptn f explrng the ray agram that apples t ths prblem. The stance between

More information

Circuits Op-Amp. Interaction of Circuit Elements. Quick Check How does closing the switch affect V o and I o?

Circuits Op-Amp. Interaction of Circuit Elements. Quick Check How does closing the switch affect V o and I o? Crcuts Op-Amp ENGG1015 1 st Semester, 01 Interactn f Crcut Elements Crcut desgn s cmplcated by nteractns amng the elements. Addng an element changes vltages & currents thrughut crcut. Example: clsng a

More information

Feedback Principle :-

Feedback Principle :- Feedback Prncple : Feedback amplfer s that n whch a part f the utput f the basc amplfer s returned back t the nput termnal and mxed up wth the nternal nput sgnal. The sub netwrks f feedback amplfer are:

More information

A method of constructing rock-analysis diagrams a statistical basks.

A method of constructing rock-analysis diagrams a statistical basks. 130 A methd f cnstructng rck-analyss dagrams a statstcal basks. 0T~ By W. ALF~.D ll~ch).ra)so.~, ~.Se., B.Se. (Eng.), F.G.S. Lecturer n Petrlgy, Unversty Cllege, Nttngham. [Read January 18, 1921.] D R.

More information

DIMENSION REDUCTION FOR CENSORED REGRESSION DATA

DIMENSION REDUCTION FOR CENSORED REGRESSION DATA The Annals f Statstcs 1999, Vl. 27, N. 1, 1 23 DIMENSION REDUCTION FOR CENSORED REGRESSION DATA BY KER-CHAU LI, 1 JANE-LING WANG 2 AND CHUN-HOUH CHEN Unversty f Calfrna, Ls Angeles, Unversty f Calfrna,

More information

A Proposal of Heating Load Calculation considering Stack Effect in High-rise Buildings

A Proposal of Heating Load Calculation considering Stack Effect in High-rise Buildings A Prpsal f Heatng Lad Calculatn cnsderng Stack Effect n Hgh-rse Buldngs *Dsam Sng 1) and Tae-Hyuk Kang 2) 1) Department f Archtectural Engneerng, Sungkyunkwan Unversty, 2066 Sebu-r, Jangan-gu, Suwn, 440-746,

More information

Chapter 11: Simple Linear Regression and Correlation

Chapter 11: Simple Linear Regression and Correlation Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests

More information

SELECTION OF MODEL PARAMETERS OF BIOGAS IC ENGINE. Karol Cupiał, Grzegorz Katolik

SELECTION OF MODEL PARAMETERS OF BIOGAS IC ENGINE. Karol Cupiał, Grzegorz Katolik TEKA Km. Mt. Energ. Rln., 2006, 6A, 32 38 SELECTION OF MODEL PARAMETERS OF BIOGAS IC ENGINE Karl Cupał, Grzegrz Katlk Insttute f Internal Cmbustn Engnes and Cntrl Engneerng Techncal Unversty f Częstchwa

More information

Exercises H /OOA> f Wo AJoTHS l^»-l S. m^ttrt /A/ ?C,0&L6M5 INFERENCE FOR DISTRIBUTIONS OF CATEGORICAL DATA. tts^e&n tai-ns 5 2%-cas-hews^, 27%

Exercises H /OOA> f Wo AJoTHS l^»-l S. m^ttrt /A/ ?C,0&L6M5 INFERENCE FOR DISTRIBUTIONS OF CATEGORICAL DATA. tts^e&n tai-ns 5 2%-cas-hews^, 27% /A/ mttrt?c,&l6m5 INFERENCE FOR DISTRIBUTIONS OF CATEGORICAL DATA Exercses, nuts! A cmpany clams that each batch f ttse&n ta-ns 5 2%-cas-hews, 27% almnds, 13% macadama nuts, and 8% brazl nuts. T test ths

More information

IGEE 401 Power Electronic Systems. Solution to Midterm Examination Fall 2004

IGEE 401 Power Electronic Systems. Solution to Midterm Examination Fall 2004 Jós, G GEE 401 wer Electrnc Systems Slutn t Mdterm Examnatn Fall 2004 Specal nstructns: - Duratn: 75 mnutes. - Materal allwed: a crb sheet (duble sded 8.5 x 11), calculatr. - Attempt all questns. Make

More information

Lucas Imperfect Information Model

Lucas Imperfect Information Model Lucas Imerfect Infrmatn Mdel 93 Lucas Imerfect Infrmatn Mdel The Lucas mdel was the frst f the mdern, mcrfundatns mdels f aggregate suly and macrecnmcs It bult drectly n the Fredman-Phels analyss f the

More information

PT326 PROCESS TRAINER

PT326 PROCESS TRAINER PT326 PROCESS TRAINER 1. Descrptn f the Apparatus PT 326 Prcess Traner The PT 326 Prcess Traner mdels cmmn ndustral stuatns n whch temperature cntrl s requred n the presence f transprt delays and transfer

More information

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank CAUSAL INFERENCE Technical Track Sessin I Phillippe Leite The Wrld Bank These slides were develped by Christel Vermeersch and mdified by Phillippe Leite fr the purpse f this wrkshp Plicy questins are causal

More information

14 The Boole/Stone algebra of sets

14 The Boole/Stone algebra of sets 14 The Ble/Stne algebra f sets 14.1. Lattces and Blean algebras. Gven a set A, the subsets f A admt the fllwng smple and famlar peratns n them: (ntersectn), (unn) and - (cmplementatn). If X, Y A, then

More information

Comparison of Building Codes and Insulation in China and Iceland

Comparison of Building Codes and Insulation in China and Iceland Prceedngs Wrld Gethermal Cngress 00 Bal, Indnesa, 5-9 prl 00 Cmparsn f Buldng Cdes and Insulatn n Chna and Iceland Hayan Le and Pall Valdmarssn Tanjn Gethermal esearch & Tranng Centre, Tanjn Unversty,

More information

CONVEX COMBINATIONS OF ANALYTIC FUNCTIONS

CONVEX COMBINATIONS OF ANALYTIC FUNCTIONS rnat. J. Math. & Math. S. Vl. 6 N. (983) 33534 335 ON THE RADUS OF UNVALENCE OF CONVEX COMBNATONS OF ANALYTC FUNCTONS KHALDA. NOOR, FATMA M. ALOBOUD and NAEELA ALDHAN Mathematcs Department Scence Cllege

More information

CTN 2/23/16. EE 247B/ME 218: Introduction to MEMS Design Lecture 11m2: Mechanics of Materials. Copyright 2016 Regents of the University of California

CTN 2/23/16. EE 247B/ME 218: Introduction to MEMS Design Lecture 11m2: Mechanics of Materials. Copyright 2016 Regents of the University of California Vlume Change fr a Unaxal Stress Istrpc lastcty n 3D Istrpc = same n all drectns The cmplete stress-stran relatns fr an strpc elastc Stresses actng n a dfferental vlume element sld n 3D: (.e., a generalzed

More information

Physic 231 Lecture 33

Physic 231 Lecture 33 Physc 231 Lecture 33 Man pnts f tday s lecture: eat and heat capacty: Q cm Phase transtns and latent heat: Q Lm ( ) eat flw Q k 2 1 t L Examples f heat cnductvty, R values fr nsulatrs Cnvectn R L / k Radatn

More information

Lead/Lag Compensator Frequency Domain Properties and Design Methods

Lead/Lag Compensator Frequency Domain Properties and Design Methods Lectures 6 and 7 Lead/Lag Cmpensatr Frequency Dmain Prperties and Design Methds Definitin Cnsider the cmpensatr (ie cntrller Fr, it is called a lag cmpensatr s K Fr s, it is called a lead cmpensatr Ntatin

More information

STAT 3008 Applied Regression Analysis

STAT 3008 Applied Regression Analysis STAT 3008 Appled Regresson Analyss Tutoral : Smple Lnear Regresson LAI Chun He Department of Statstcs, The Chnese Unversty of Hong Kong 1 Model Assumpton To quantfy the relatonshp between two factors,

More information

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution Department of Statstcs Unversty of Toronto STA35HS / HS Desgn and Analyss of Experments Term Test - Wnter - Soluton February, Last Name: Frst Name: Student Number: Instructons: Tme: hours. Ads: a non-programmable

More information

Problem Set 5 Solutions - McQuarrie Problems 3.20 MIT Dr. Anton Van Der Ven

Problem Set 5 Solutions - McQuarrie Problems 3.20 MIT Dr. Anton Van Der Ven Prblem Set 5 Slutns - McQuarre Prblems 3.0 MIT Dr. Antn Van Der Ven Fall Fall 003 001 Prblem 3-4 We have t derve the thermdynamc prpertes f an deal mnatmc gas frm the fllwng: = e q 3 m = e and q = V s

More information

, which yields. where z1. and z2

, which yields. where z1. and z2 The Gaussian r Nrmal PDF, Page 1 The Gaussian r Nrmal Prbability Density Functin Authr: Jhn M Cimbala, Penn State University Latest revisin: 11 September 13 The Gaussian r Nrmal Prbability Density Functin

More information

Mode-Frequency Analysis of Laminated Spherical Shell

Mode-Frequency Analysis of Laminated Spherical Shell Mde-Frequency Analyss f Lamnated Sphercal Shell Umut Tpal Department f Cvl Engneerng Karadenz Techncal Unversty 080, Trabzn, Turkey umut@ktu.edu.tr Sessn ENG P50-00 Abstract Ths paper deals wth mde-frequency

More information

Exploiting vector space properties for the global optimization of process networks

Exploiting vector space properties for the global optimization of process networks Exptng vectr space prpertes fr the gbal ptmzatn f prcess netwrks Juan ab Ruz Ignac Grssmann Enterprse Wde Optmzatn Meetng March 00 Mtvatn - The ptmzatn f prcess netwrks s ne f the mst frequent prblems

More information

A Note on Equivalences in Measuring Returns to Scale

A Note on Equivalences in Measuring Returns to Scale Internatnal Jurnal f Busness and Ecnmcs, 2013, Vl. 12, N. 1, 85-89 A Nte n Equvalences n Measurng Returns t Scale Valentn Zelenuk Schl f Ecnmcs and Centre fr Effcenc and Prductvt Analss, The Unverst f

More information

Chapter 6 : Gibbs Free Energy

Chapter 6 : Gibbs Free Energy Wnter 01 Chem 54: ntrductry hermdynamcs Chapter 6 : Gbbs Free Energy... 64 Defntn f G, A... 64 Mawell Relatns... 65 Gbbs Free Energy G(,) (ure substances)... 67 Gbbs Free Energy fr Mtures... 68 ΔG f deal

More information

A/2 l,k. Problem 1 STRATEGY. KNOWN Resistance of a complete spherical shell: r rk. Inner and outer radii

A/2 l,k. Problem 1 STRATEGY. KNOWN Resistance of a complete spherical shell: r rk. Inner and outer radii Prblem 1 STRATEGY KNOWN Resstance f a cmplete sphercal shell: R ( r r / (4 π r rk sphere Inner an uter ra r an r, SOLUTION Part 1: Resstance f a hemsphercal shell: T calculate the resstance f the hemsphere,

More information

Linear Plus Linear Fractional Capacitated Transportation Problem with Restricted Flow

Linear Plus Linear Fractional Capacitated Transportation Problem with Restricted Flow Amercan urnal f Operatns Research,,, 58-588 Publshed Onlne Nvember (http://www.scrp.rg/urnal/ar) http://dx.d.rg/.46/ar..655 Lnear Plus Lnear Fractnal Capactated Transprtatn Prblem wth Restrcted Flw Kavta

More information

Statistics for Economics & Business

Statistics for Economics & Business Statstcs for Economcs & Busness Smple Lnear Regresson Learnng Objectves In ths chapter, you learn: How to use regresson analyss to predct the value of a dependent varable based on an ndependent varable

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Design of Analog Integrated Circuits

Design of Analog Integrated Circuits Desgn f Analg Integrated Crcuts I. Amplfers Desgn f Analg Integrated Crcuts Fall 2012, Dr. Guxng Wang 1 Oerew Basc MOS amplfer structures Cmmn-Surce Amplfer Surce Fllwer Cmmn-Gate Amplfer Desgn f Analg

More information

Chem 204A, Fall 2004, Mid-term (II)

Chem 204A, Fall 2004, Mid-term (II) Frst tw letters f yur last name Last ame Frst ame McGll ID Chem 204A, Fall 2004, Md-term (II) Read these nstructns carefully befre yu start tal me: 2 hurs 50 mnutes (6:05 PM 8:55 PM) 1. hs exam has ttal

More information

Tubular Flow with Laminar Flow (CHE 512) M.P. Dudukovic Chemical Reaction Engineering Laboratory (CREL), Washington University, St.

Tubular Flow with Laminar Flow (CHE 512) M.P. Dudukovic Chemical Reaction Engineering Laboratory (CREL), Washington University, St. Tubular Flw wth Lamnar Flw (CHE 5) M.P. Dudukvc Chemcal Reactn Engneerng Labratry (CREL), Washngtn Unversty, St. Lus, MO 4. TUBULAR REACTORS WITH LAMINAR FLOW Tubular reactrs n whch hmgeneus reactns are

More information

How do scientists measure trees? What is DBH?

How do scientists measure trees? What is DBH? Hw d scientists measure trees? What is DBH? Purpse Students develp an understanding f tree size and hw scientists measure trees. Students bserve and measure tree ckies and explre the relatinship between

More information

SCALE MIXTURES DISTRIBUTIONS IN INSURANCE APPLICATIONS ABSTRACT KEYWORDS

SCALE MIXTURES DISTRIBUTIONS IN INSURANCE APPLICATIONS ABSTRACT KEYWORDS SCALE MIXTURES DISTRIBUTINS IN INSURANCE ALICATINS BY S.T. BRIS CHY AND C.M. CHAN ABSTRACT In ths paper nn-nrmal dstrbutns va scale mxtures are ntrduced nt nsurance applcatns. The symmetrc dstrbutns f

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

ME2142/ME2142E Feedback Control Systems. Modelling of Physical Systems The Transfer Function

ME2142/ME2142E Feedback Control Systems. Modelling of Physical Systems The Transfer Function Mdellng Physcal Systems The Transer Functn Derental Equatns U Plant Y In the plant shwn, the nput u aects the respnse the utput y. In general, the dynamcs ths respnse can be descrbed by a derental equatn

More information

Basics of heteroskedasticity

Basics of heteroskedasticity Sect 8 Heterskedastcty ascs f heterskedastcty We have assumed up t w ( ur SR ad MR assumpts) that the varace f the errr term was cstat acrss bservats Ths s urealstc may r mst ecmetrc applcats, especally

More information

55:041 Electronic Circuits

55:041 Electronic Circuits 55:04 Electrnc Crcuts Feedback & Stablty Sectns f Chapter 2. Kruger Feedback & Stablty Cnfguratn f Feedback mplfer S S S S fb Negate feedback S S S fb S S S S S β s the feedback transfer functn Implct

More information

MAXIMIN CLUSTERS FOR NEAR-REPLICATE REGRESSION LACK OF FIT TESTS. BY FORREST R. MILLER, JAMES W. NEILL AND BRIAN W. SHERFEY Kansas State University

MAXIMIN CLUSTERS FOR NEAR-REPLICATE REGRESSION LACK OF FIT TESTS. BY FORREST R. MILLER, JAMES W. NEILL AND BRIAN W. SHERFEY Kansas State University The Annals f Statstcs 1998, Vl. 6, N. 4, 14111433 MAXIMIN CLUSTERS FOR NEAR-REPLICATE REGRESSION LACK OF FIT TESTS BY FORREST R. MILLER, JAMES W. NEILL AND BRIAN W. SHERFEY Kansas State Unversty T assess

More information

_J _J J J J J J J J _. 7 particles in the blue state; 3 particles in the red state: 720 configurations _J J J _J J J J J J J J _

_J _J J J J J J J J _. 7 particles in the blue state; 3 particles in the red state: 720 configurations _J J J _J J J J J J J J _ Dsrder and Suppse I have 10 partcles that can be n ne f tw states ether the blue state r the red state. Hw many dfferent ways can we arrange thse partcles amng the states? All partcles n the blue state:

More information

Feature Selection for Time Series Modeling *

Feature Selection for Time Series Modeling * Jurnal f Intellgent Learnng Systems Applcatns, 013, 5, 15-164 http://dxdrg/10436/lsa01353017 Publshed Onlne August 013 (http://wwwscrprg/urnal/lsa) Feature Selectn fr Tme Seres Mdelng * Qng-Gu Wang, Xan

More information

Differentiation Applications 1: Related Rates

Differentiation Applications 1: Related Rates Differentiatin Applicatins 1: Related Rates 151 Differentiatin Applicatins 1: Related Rates Mdel 1: Sliding Ladder 10 ladder y 10 ladder 10 ladder A 10 ft ladder is leaning against a wall when the bttm

More information

Department of Quantitative Methods & Information Systems. Time Series and Their Components QMIS 320. Chapter 6

Department of Quantitative Methods & Information Systems. Time Series and Their Components QMIS 320. Chapter 6 Department of Quanttatve Methods & Informaton Systems Tme Seres and Ther Components QMIS 30 Chapter 6 Fall 00 Dr. Mohammad Zanal These sldes were modfed from ther orgnal source for educatonal purpose only.

More information

Van der Waals-coupled electronic states in incommensurate double-walled carbon nanotubes

Van der Waals-coupled electronic states in incommensurate double-walled carbon nanotubes Kahu Lu* 1, Chenha Jn* 1, Xapng Hng 1, Jhn Km 1, Alex Zettl 1,2, Enge Wang 3, Feng Wang 1,2 Van der Waals-cupled electrnc states n ncmmensurate duble-walled carbn nantubes S1. Smulated absrptn spectra

More information

Physics 2B Chapter 23 Notes - Faraday s Law & Inductors Spring 2018

Physics 2B Chapter 23 Notes - Faraday s Law & Inductors Spring 2018 Michael Faraday lived in the Lndn area frm 1791 t 1867. He was 29 years ld when Hand Oersted, in 1820, accidentally discvered that electric current creates magnetic field. Thrugh empirical bservatin and

More information

Where a licence is displayed above, please note the terms and conditions of the licence govern your use of this document.

Where a licence is displayed above, please note the terms and conditions of the licence govern your use of this document. Mdellng the seepage flw durng cassn nstallatn n a natural seabed Faramarz, Asaad; Faz, Khyar; Drar, Samr; Mehravar, Mura; Harreche, Ouahd Dcument Versn Publsher's PDF, als knwn as Versn f recrd Ctatn fr

More information

Approach: (Equilibrium) TD analysis, i.e., conservation eqns., state equations Issues: how to deal with

Approach: (Equilibrium) TD analysis, i.e., conservation eqns., state equations Issues: how to deal with Schl f Aerspace Chemcal D: Mtvatn Prevus D Analyss cnsdered systems where cmpstn f flud was frzen fxed chemcal cmpstn Chemcally eactng Flw but there are numerus stuatns n prpulsn systems where chemcal

More information

Modelling of Clock Behaviour. Don Percival. Applied Physics Laboratory University of Washington Seattle, Washington, USA

Modelling of Clock Behaviour. Don Percival. Applied Physics Laboratory University of Washington Seattle, Washington, USA Mdelling f Clck Behaviur Dn Percival Applied Physics Labratry University f Washingtn Seattle, Washingtn, USA verheads and paper fr talk available at http://faculty.washingtn.edu/dbp/talks.html 1 Overview

More information

t r m o o H Is The Sensitive Information Of Your Company Completely Secure?

t r m o o H Is The Sensitive Information Of Your Company Completely Secure? : n t a c f t r e C 1 0 0 7 l 2 l O W S I y n a p m C r u Y w H t f e n e B Cyber crmnals are fndng ncreasngly clever ways every day t be able t peek ver yur shulder, and wth ths llegal ndustry beng an

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Negative Binomial Regression

Negative Binomial Regression STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...

More information

UNIVERSITY OF TORONTO Faculty of Arts and Science. December 2005 Examinations STA437H1F/STA1005HF. Duration - 3 hours

UNIVERSITY OF TORONTO Faculty of Arts and Science. December 2005 Examinations STA437H1F/STA1005HF. Duration - 3 hours UNIVERSITY OF TORONTO Faculty of Arts and Scence December 005 Examnatons STA47HF/STA005HF Duraton - hours AIDS ALLOWED: (to be suppled by the student) Non-programmable calculator One handwrtten 8.5'' x

More information

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y)

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y) Secton 1.5 Correlaton In the prevous sectons, we looked at regresson and the value r was a measurement of how much of the varaton n y can be attrbuted to the lnear relatonshp between y and x. In ths secton,

More information

AP Statistics Notes Unit Two: The Normal Distributions

AP Statistics Notes Unit Two: The Normal Distributions AP Statistics Ntes Unit Tw: The Nrmal Distributins Syllabus Objectives: 1.5 The student will summarize distributins f data measuring the psitin using quartiles, percentiles, and standardized scres (z-scres).

More information

Economics 130. Lecture 4 Simple Linear Regression Continued

Economics 130. Lecture 4 Simple Linear Regression Continued Economcs 130 Lecture 4 Contnued Readngs for Week 4 Text, Chapter and 3. We contnue wth addressng our second ssue + add n how we evaluate these relatonshps: Where do we get data to do ths analyss? How do

More information

Monin Obukhov Similarity and Local-Free-Convection Scaling in the Atmospheric Boundary Layer Using Matched Asymptotic Expansions

Monin Obukhov Similarity and Local-Free-Convection Scaling in the Atmospheric Boundary Layer Using Matched Asymptotic Expansions OCTOBER 08 T O N G A N D D I N G 369 Mnn Obukhv Smlarty cal-free-cnvectn Scalng n the Atmspherc Bundary ayer Usng Matched Asympttc Expansns CHENNING TONG AND MENGJIE DING Department f Mechancal Engneerng

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

We can see from the graph above that the intersection is, i.e., [ ).

We can see from the graph above that the intersection is, i.e., [ ). MTH 111 Cllege Algebra Lecture Ntes July 2, 2014 Functin Arithmetic: With nt t much difficulty, we ntice that inputs f functins are numbers, and utputs f functins are numbers. S whatever we can d with

More information

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands Content. Inference on Regresson Parameters a. Fndng Mean, s.d and covarance amongst estmates.. Confdence Intervals and Workng Hotellng Bands 3. Cochran s Theorem 4. General Lnear Testng 5. Measures of

More information

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS M. Krshna Reddy, B. Naveen Kumar and Y. Ramu Department of Statstcs, Osmana Unversty, Hyderabad -500 007, Inda. nanbyrozu@gmal.com, ramu0@gmal.com

More information

A Matrix Representation of Panel Data

A Matrix Representation of Panel Data web Extensin 6 Appendix 6.A A Matrix Representatin f Panel Data Panel data mdels cme in tw brad varieties, distinct intercept DGPs and errr cmpnent DGPs. his appendix presents matrix algebra representatins

More information

Big Data Analytics! Special Topics for Computer Science CSE CSE Mar 31

Big Data Analytics! Special Topics for Computer Science CSE CSE Mar 31 Bg Data Analytcs! Specal Tpcs fr Cmputer Scence CSE 4095-001 CSE 5095-005! Mar 31 Fe Wang Asscate Prfessr Department f Cmputer Scence and Engneerng fe_wang@ucnn.edu Intrductn t Deep Learnng Perceptrn In

More information

III. Operational Amplifiers

III. Operational Amplifiers III. Operatnal Amplfers Amplfers are tw-prt netwrks n whch the utput vltage r current s drectly prprtnal t ether nput vltage r current. Fur dfferent knds f amplfers ext: ltage amplfer: Current amplfer:

More information

The Ordinary Least Squares (OLS) Estimator

The Ordinary Least Squares (OLS) Estimator The Ordnary Least Squares (OLS) Estmator 1 Regresson Analyss Regresson Analyss: a statstcal technque for nvestgatng and modelng the relatonshp between varables. Applcatons: Engneerng, the physcal and chemcal

More information

Comparison of Regression Lines

Comparison of Regression Lines STATGRAPHICS Rev. 9/13/2013 Comparson of Regresson Lnes Summary... 1 Data Input... 3 Analyss Summary... 4 Plot of Ftted Model... 6 Condtonal Sums of Squares... 6 Analyss Optons... 7 Forecasts... 8 Confdence

More information

Differential Fault Analysis of SHA3-224 and SHA3-256

Differential Fault Analysis of SHA3-224 and SHA3-256 Dfferental Fault Analyss f SHA3-4 and SHA3-256 Pe Lu, Yuns Fe, Lwe Zhang, and A Adam Dng slencelu@gmalcm, yfe@eceneuedu, mathlwe@gmalcm, adng@neuedu Electrcal & Cmputer Engneerng Department, Nrtheastern

More information

Interference is when two (or more) sets of waves meet and combine to produce a new pattern.

Interference is when two (or more) sets of waves meet and combine to produce a new pattern. Interference Interference is when tw (r mre) sets f waves meet and cmbine t prduce a new pattern. This pattern can vary depending n the riginal wave directin, wavelength, amplitude, etc. The tw mst extreme

More information

x yi In chapter 14, we want to perform inference (i.e. calculate confidence intervals and perform tests of significance) in this setting.

x yi In chapter 14, we want to perform inference (i.e. calculate confidence intervals and perform tests of significance) in this setting. The Practce of Statstcs, nd ed. Chapter 14 Inference for Regresson Introducton In chapter 3 we used a least-squares regresson lne (LSRL) to represent a lnear relatonshp etween two quanttatve explanator

More information