WPCQ
2
BJZCourier3|j
w RomanTimes New Roman BoldX@HP LaserJet 5SiHPLAS5SI.PRSx
@\&YX@CourierTimes New Roman#X\ P6G;ɒP#X01Í ÍX01Í Í#Xj\ P6G;ynXP#2B<KTDKCourierTimes New Roman"i~'^09CSS999S]+9+/SSSSSSSSSS//]]]Ixnnxg]xx9?xgxx]xn]gxxxxg9/9MS9ISISI9SS//S/SSSS9?/SSxSSIP!PZ9+ZM999+999999S9S/xIxIxIxIxIlnIgIgIgIgI9/9/9/9/xSxSxSxSxSxSxSxSxSxSxIxSxRxSxSxS]SxIxIxInInInZnIxigIgIgIgIxSxSxSxZxSxZxS9/9S999Su]ZZxSg/gCg9g9g/xSbxSxSxSxSxn9n9n9]?]?]?]ZgFg/gMxSxSxSxSxSxSxxZgIgIgIxSg9xS]?g9xSi+SS88WuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxN8HH"&H>XHH8HB8>HH^HH>"".2",2,2,"222N2222"&22H22,006"6."""""""""2"2H,H,H,H,H,XAB,>,>,>,>,""""H2H2H2H2H2H2H2H2H2H2H,H2H1H2H2H282H,H,H,B,B,B6B,H?>,>,>,>,H2H2H2H6H2H6H2""2"""2F866H2>>(>">">H2;H2H2H2H2XHB"B"B"8&8&8&86>*>>.H2H2H2H2H2H2^HH6>,>,>,H2>"H28&>"H2?22!!WFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxN$<<$.2",2222`2 LL2 LL2L"",,2d""(-((((((((((---#J:55:2-:::2F::-:5-2::K::2%(#(#(#(((>((((((:((#&&++%((:#:#:#:#:#F45#2#2#2#2#:(:(:(:(:(:(:(:(:(:(:#:(:':(:(:(-(:#:#:#5#5#5+5#:22#2#2#2#:(:(:(:+:(:+:(((8-++:(22 222:(/:(:(:(:(F:555----+2"22%:(:(:(:(:(:(K::+2#2#2#:(2:(-2:(2((W888888888888888888888888888888888888888888888888xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxN00%(#((((M(==(==(=##(P0P((N0=PP00/BB--P#(CP"5555==JPP(=P0.+(-N00P(2KK+\vT"i~'^:DTddDDDd4D48ddddddddddDDd||||DXp||dp||ppL8LTdDddXdX8dd88X8ddddLL8dXXXLP8PlD4lTDDD4DDDDDDdDd8|d|d|d|d|dX|X|X|X|XD8D8D8D8dddddddddpX|ddddpXd|d|d|d|dXXlXx|X|X|X|XdddldldD8DdDDDddllXp8pHpDp@p8dtdddd|L|L|LdLdLdLllpHp8pTddddddplpLpLpLdpDddLpDpdx4ddC,CWddddddddddddddddddddddddddddddddddddddddNHxxHhdLdddddd8@d<@d<DDppdDDxddxHxxHkddDpd<"dxtldxxd"i~'^#)0<55).
Based upon this evidence Sprint suggests that the data in the Commissions proposed data set
or the proposed regression equation appears to be "severely tainted" and recommends
dismissing all conclusions suggested as a result of this tainted data set and misspecified
regression model.
Xm47.` ` We reject Sprints argument. While we acknowledge that there is collinearity
amongst the explanatory variables, we note that this is typically the case in multiple regression
models. Anderson, Sweeney, and Williams note that 8most independent variables in a
X(4multiple regression problem are correlated to some degree with one another.(Z{
{O3'ԍ See David Anderson, Dennis Sweeney, and Thomas Williams (1996), Statistics for Business and
{O'Economics, Sixth Edition at page 597. Similarly
Fomby, Hill, and Johnson note, [f]requently in nonexperimental situations, some explanatory
variables exhibit little variation, or the variation they do exhibit is systematically related to
X4variation in other explanatory variables.{
{OJ'ԍ See Thomas Fomby, R. Carter Hill, and Stanley Johnson (1988), Advanced Econometric Methods at page
283.
Muliticollinearity does not as Sprint implies indicate that the regression model is mis
X4specified.{
{O_"'ԍ See William Green (1990), Econometric Analysis at 278 (noting that the case of near collinearity or high
intercorrelation among the variables is 8 a statistical problem. The difficulty in estimation is not one of
identification but of precision.), or Fomby, Hill, and Johnson at 284 (noting that the primary statistical
consequence of multicollinearity is that one or more of the estimated coefficients of the linear model may have
large standard errors.) Therefore, the issues of misspecification and mutlticollinearity are independent,
and Sprint provides no evidence that the regression model is misspecified.
"p0*(("Ԍ
X48.` ` Even with multicollinearity the least squares estimate is the minimum variance
linear unbiased estimator, its standard error is correct, and the conventional confidence
X4interval and hypothesis tests are valid.t{
{OK'ԍ See Arthur Goldberger (1991), A Course in Econometrics at 246.t The least squares estimates and hence forecasts based
upon them are also best linear unbiased estimates and maximum likelihood estimates and
X4hence are unbiased, efficient, and consistent.rZ{
{O'ԍ See Ramu Ramanthan (1989), Introductory Econometrics at 232.r Furthermore Ramanthan notes that
Multicollinearity may not affect the forecasting performance of a model and may possibly
Xv4even improve it.rv{
{O'ԍ See Ramu Ramanthan (1989), Introductory Econometrics at 233.r
XH49.` ` Sprint also raises concerns in their comments regarding the lack of statistical
X14significance of individual parameters in our estimates.f1~{
{O`'ԍ See Sprint Inputs Further Notice Comments at 44.f However as Golberger notes, while
muliticollinearity may make the estimates of individual parameters less precise, it may
X
4 facilitate the precise estimation of particular combinations of elements.t
{
{O'ԍ See Arthur Goldberger (1991), A Course in Econometrics at 250.t#5\ P6QمP# #XP\ P6QynXP#For #XP\ P6QynXP#example Sprint
expresses concern that the lines variable "by itself" should be more significant. Staff analysis
indicates, however, that jointly the variables Lines and Lines/Time are statistically significant,
indicating that switches increase significantly in cost when additional lines are purchased at
installation.. Therefore, one would be in error to conclude that, based upon individual t
X4statistics, switch costs do not vary with line size. #XP\ P6QynXP#