From 4e55846db8cdfbb55bb7278c1eecb330ad0d9f43 Mon Sep 17 00:00:00 2001 From: Michael Friendly Date: Sat, 23 Nov 2024 18:29:16 -0500 Subject: [PATCH] edits to Ch08 re MT comments --- 08-collinearity-ridge.qmd | 41 ++++--- 10-mlm-review.qmd | 8 +- 11-mlm-viz.qmd | 22 +++- R/arcmanov.sas | 181 +++++++++++++++++++++++++++++ R/cars-colldiag.R | 22 ++++ R/collin-data-beta.R | 4 +- bib/pkgs.txt | 10 ++ images/arcmanov1.png | Bin 0 -> 26900 bytes images/arcmanov2.png | Bin 0 -> 29362 bytes working-text/MichaelT_comments.Rmd | 1 + 10 files changed, 269 insertions(+), 20 deletions(-) create mode 100644 R/arcmanov.sas create mode 100644 images/arcmanov1.png create mode 100644 images/arcmanov2.png diff --git a/08-collinearity-ridge.qmd b/08-collinearity-ridge.qmd index 67f98dbf..45cf6004 100644 --- a/08-collinearity-ridge.qmd +++ b/08-collinearity-ridge.qmd @@ -25,7 +25,7 @@ and presents some novel visual tools for these purposes using the `r pkg("VisCo One class of solutions for collinearity involves _regularization methods_ such as ridge regression. Another collection of graphical methods, generalized ridge trace plots, implemented -in the `r pkg("genridge")`` package, sheds further light on what is accomplished by this technique. +in the `r pkg("genridge")` package, sheds further light on what is accomplished by this technique. More generally, the methods of this chapter are further examples of how data and confidence ellipsoids can be used to visualize bias **and** precision of regression estimates. @@ -109,8 +109,8 @@ The sampling variances and covariances of the estimated coefficients is $\text{Var} (\widehat{\mathbf{b}}) = \sigma_\epsilon^2 \times (\mathbf{X}^\mathsf{T} \mathbf{X})^{-1}$ and $\sigma_\epsilon^2$ is the variance of the residuals $\mathbf{\epsilon}$, estimated by the mean squared error (MSE). -In the limiting case, when one $x_i$ is _perfectly_ -predictable from the other $x$s, i.e., $R^2 (x_i | \text{other }x) = 1$, + +In the limiting case, collinearity becomes particularly problematic when one $x_i$ is _perfectly_ predictable from the other $x$s, i.e., $R^2 (x_i | \text{other }x) = 1$. This is problematic because: * there is no _unique_ solution for the regression coefficients $\mathbf{b} = (\mathbf{X}^\mathsf{T} \mathbf{X})^{-1} \mathbf{X} \mathbf{y}$; @@ -121,10 +121,13 @@ This extreme case reflects a situation when one or more predictors are effective when you include two variables $x$ and $y$ and their sum $z = x + y$ in a model. For instance, a dataset may include variables for income, expenses, and savings. But income is the sum of expenses and savings, so not all three should be used as predictors. -A more subtly case is the use _ipsatized_, defined as + +A more subtle case is the use _ipsatized_, defined as scores that sum to a constant, such as proportions of a total. You might have scores on tests of reading, math, spelling and geography. With ipsatized scores, any one of these -is necessarily 1 $-$ sum of the others. +is necessarily 1 $-$ sum of the others, i.e., if reading is 0.5, math and geography are +both 0.15, then geography must be 0.2. Once thre of the four scores are known, the last +provides no new information. More generally, collinearity refers to the case when there are very high **multiple correlations** among the predictors, such as $R^2 (x_i | \text{other }x) \ge 0.9$. @@ -141,7 +144,7 @@ so a coefficient that would significant if the predictors were uncorrelated beco when collinearity is present. * Thus you may find a situation where an overall model is highly significant (large $F$-statistic), while no (or few) of the individual predictors are. This is a puzzlement! -* Beyond this, the least squares solution may have poor numerical accurracy [@Longley:1967], because +* Beyond this, the least squares solution may have poor numerical accuracy [@Longley:1967], because the solution depends inversely on the determinant $|\,\mathbf{X}^\mathsf{T} \mathbf{X}\,|$, which approaches 0 as multiple correlations increase. * There is an interpretive problem as well. Recall that the coefficients $\hat{b}$ are _partial coefficients_, meaning that they estimate change $\Delta y$ in $y$ when $x$ changes by one unit $\Delta x$, but holding **all other variables @@ -216,8 +219,8 @@ beta <- c(3, 3) # true coefficients # Specify a covariance matrix, with standard deviations # s[1], s[2] and correlation r Cov <- function(s, r){ - matrix(c(s[1], r * prod(s), - r * prod(s), s[2]), nrow = 2, ncol = 2) + matrix(c(s[1], r * s[1]*s[2], + r * s[1]*s[2], s[2]), nrow = 2, ncol = 2) } # Generate a dataframe of X, y for each rho @@ -270,8 +273,8 @@ do_plots <- function(XY, mod, r) { confidenceEllipse(mod, col = "red", fill = TRUE, fill.alpha = 0.1, - xlab = "x1 coefficient", - ylab = "x2 coefficient", + xlab = expression(paste("x1 coefficient, ", beta[1])), + ylab = expression(paste("x2 coefficient, ", beta[2])), xlim = c(-5, 10), ylim = c(-5, 10), asp = 1) @@ -360,7 +363,7 @@ of such terms, relative to what would be obtained with uncorrelated data. Visual by comparing the areas of the ellipses in the bottom row of @fig-collin-data-beta. Because the magnitude of the GVIF increases with the number of degrees of freedom for the set of parameters, Fox & Monette suggest the analog $\sqrt{\text{GVIF}^{1/2 \text{df}}}$ as the measure of impact on standard -errors. +errors. This is what `car::vif()` calculates for a factor or other term with more than 1 df. **Example**: This example uses the `cars` dataset in the `VisCollin` package containing various measures of size and performance on 406 models of automobiles from 1982. Interest is focused on predicting gas mileage, `mpg`. @@ -375,11 +378,13 @@ accelerate from 0 -- 60 mph and model year (1970--1982). Perhaps surprisingly, o significantly predict gas mileage. What's going on here? ```{r cars-mod} -cars.mod <- lm (mpg ~ cylinder + engine + horse + weight + accel + year, +cars.mod <- lm (mpg ~ cylinder + engine + horse + + weight + accel + year, data=cars) Anova(cars.mod) ``` + We check the variance inflation factors, using `car::vif()`. We see that most predictors have very high VIFs, indicating moderately severe multicollinearity. @@ -390,10 +395,20 @@ sqrt(vif(cars.mod)) ``` According to $\sqrt{\text{VIF}}$, the standard error of `cylinder` has been -multiplied by 3.26 and it's $t$-value divided by this number, +multiplied by $\sqrt{10.63} = 3.26$ and it's $t$-value is divided by this number, compared with the case when all predictors are uncorrelated. `engine`, `horse` and `weight` suffer a similar fate. +If we also included the factor `origin` in the models, we would get the generalized GVIF: + +```{r cars-mod2} +cars.mod2 <- lm (mpg ~ cylinder + engine + horse + + weight + accel + year + origin + data=cars) +vif(cars.mod2) +``` + + ::: {.callout-tip title="Connection with inverse of correlation matrix"} In the linear regression model with standardized predictors, diff --git a/10-mlm-review.qmd b/10-mlm-review.qmd index 70dd7139..432a65dd 100644 --- a/10-mlm-review.qmd +++ b/10-mlm-review.qmd @@ -673,7 +673,10 @@ Then, we can illustrate @eq-H-contrasts by extracting the 1 df $\mathbf{H}$ mat from the results of `linearHypothesis`. -```{r dogfood-Eqn, results='asis'} +```{r dogfood-Eqn} +#| results: asis +#| echo: false +#| eval: false options(print.latexMatrix = list(display.labels=FALSE)) SSP_H1 <- H1$SSPH |> round(digits=2) SSP_H2 <- H2$SSPH |> round(digits=2) @@ -814,7 +817,8 @@ In @fig-parenting-boxpl, I've also plotted the group means with white dots. #| code-summary: "See the ggplot code" #| fig-cap: Faceted boxplots of scores on the three parenting scales, showing also the mean for each. parenting_long <- Parenting |> - tidyr::pivot_longer(cols=caring:play, names_to = "variable") + tidyr::pivot_longer(cols=caring:play, + names_to = "variable") ggplot(parenting_long, aes(x=group, y=value, fill=group)) + diff --git a/11-mlm-viz.qmd b/11-mlm-viz.qmd index 8e610b95..539ab1c2 100644 --- a/11-mlm-viz.qmd +++ b/11-mlm-viz.qmd @@ -28,7 +28,7 @@ library(tidyr) ## HE plot framework {#sec-he-framework} @sec-Hotelling illustrated the basic ideas of the framework for visualizing multivariate -linear models in the context of a simple two group design, using Hotelling's $T^2$. These are illustrated in @fig-HE-framework. +linear models in the context of a simple two group design, using Hotelling's $T^2$. The main ideas are illustrated in @fig-HE-framework. * In data space, each group is summarized by its **data ellipse**, representing the means and covariances. @@ -36,7 +36,9 @@ linear models in the context of a simple two group design, using Hotelling's $T^ representing the pooled within-group covariance matrix, $\mathbf{S}_p$ and the data ellipse of the residuals from the model. * The MANOVA (or Hotelling's $T^2$) is formally equivalent to a **discriminant analysis**, predicting -group membership from the response variables which can be seen in data space. +group membership from the response variables which can be seen in data space. (The main difference is +emphasis and goals: MANOVA seeks to test differences among group means, while discriminant analysis aims +at classification of the observations into groups.) * This effectively projects the $p$-dimensional space of the predictors into the smaller **canonical space** that shows the greatest differences among @@ -47,10 +49,24 @@ the groups. #| echo: false #| fig-align: center #| out-width: "100%" -#| fig-cap: "The Hypothesis Error plot framework. _Source_: author" +#| fig-cap: "The Hypothesis Error plot framework for a two-group design. Above: Data ellipses can be summarized in an HE plot showing the pooled within-group error ($\\mathbf{E}$) ellipse and the $\\mathbf{H}$ "ellipse" for the group means. +#| Below: Observations projected on the line joining the means give discriminant scores which correpond to a one-dimensional canonical space." knitr::include_graphics("images/HE-framework.png") ``` +Having described the statistical ideas behind the MLM in @sec-mlm-review, we can proceed to +extend this framework for visualization ... + +```{r} +#| label: fig-arcmanov1 +#| echo: false +#| fig-align: center +#| out-width: "100%" +#| fig-cap: "Conceptual plots showing the essential ideas behind multivariate tests, in terms of the hypothesis +#| ($\\mathbf{H}$) and error ($\\mathbf{E}$) matrices for a 1-way MANOVA design with two response variables, $Y_1$ and $Y_2$" +knitr::include_graphics("images/arcmanov.png") +``` + For more complex models such as MANOVA with multiple factors or multivariate multivariate regression, there is one $\mathbf{H}$ ellipse for each term in the model. ... diff --git a/R/arcmanov.sas b/R/arcmanov.sas new file mode 100644 index 00000000..a5175ca0 --- /dev/null +++ b/R/arcmanov.sas @@ -0,0 +1,181 @@ +%include goptions; +*include arc ; +options ls=95; +goptions htext=1.7 htitle=2.5; +%let ht=1.7; +*goptions hsize=7. vsize=7.5; +Data means; + Input Group Y1 Y2; + R = .4 + .5*uniform(123131); +cards; +1 8 10 +2 10 5 +3 12 14 +4 17 20 +5 18 11 +6 20 16 +7 25 20 +8 27 26 +; +%macro fig1; +Data Fig1; + *annocms; + %annomac; + %dclanno; + %system(2,2,4); + Set means; + If _N_=1 then do; + %label(2,36.5,'Scatter around group means ',BLACK,0,0, &ht ,,6); + %label(2,35,'represented by each ellipse',BLACK,0,0,&ht,,6); + %label(38,2,'(a)',BLACK, 0,0,&ht,,4) + End; + * color = 'H' || put(Group*45,HEX3.); + color = 'BLACK'; + RX=3.3; RY=3*R; + tilt=35 + 10*uniform(0); + %ellipse(Y1,Y2,RX,RY,tilt,color=black,line=1+_N_,width=2); + %label(Y1+.5,Y2,put(_N_,2.),black,0,0,1.5,,6); +Title '(a) Individual group scatter'; +symbol1 v=dot h=1.5 i=none c=black r=8; + +proc gplot data=means; + plot y2 * y1=group / nolegend frame anno=fig1 + vaxis=axis1 haxis=axis2 hm=1 vm=1; + + axis1 order=(0 to 40 by 10) /* length=6.5 in */; + axis2 order=(0 to 40 by 10) /* length=6.5 in */; + +Proc Summary data=means; + Class GROUP; + Var Y1 Y2; + Output out=GRAND mean=Y1 Y2; +Proc Print data=GRAND; +%mend; + +%macro fig2; +Data Fig2; + %dclanno; + %system(2,2,4); + Length TEXT $30; + Set GRAND; + drop a rad xp yp Y1 Y2 group _type_ _freq_ rot tilt; + If _TYPE_=0; + %ellipse(Y1,Y2,16, 8, 50,color=BLACK,line=1,width=2); + %ellipse(Y1,Y2, 4,2.5,37,color=BLACK,line=1,width=2); + X=Y1; Y=Y2; SIZE=1.8; TEXT='+'; + FUNCTION='SYMBOL'; output; + rad = arcos(-1)/180; + Do a = 0 to 270 by 90; + ang = 37+a; + XP = Y1 + 6*cos(rad*ang); + YP = Y2 + 6*sin(rad*ang); + %line(Y1,Y2, XP,YP,*,2,1); + End; +%label(26,29,'H matrix',BLACK,0,0,&ht,,6); +%label(21,15,'E matrix',BLACK,0,0,&ht,,6); +%label(1,37.5,'Deviations of group means from',BLACK,0,0,&ht,,6); +%label(1,36.0,'grand mean (outer) and pooled ',BLACK,0,0,&ht,,6); +%label(1,34.5,'within-group (inner) ellipses.',BLACK,0,0,&ht,,6); +%label(39,37,'How big is H relative to E? ',BLACK,0,0,2,,4); +%label(38,2,'(b)',BLACK, 0,0, &ht,,4) +Title '(b) Between and Within Scatter'; +Proc print; + +Proc GPLOT data=means; + Plot Y2 * Y1 / anno=Fig2 frame + vaxis=axis1 haxis=axis2 vminor=1 hminor=1; + axis1 order=(0 to 40 by 10); + axis2 order=(0 to 40 by 10); + run; quit; +%mend; + +%macro fig3; +Data Can; + Label CAN1='First Canonical dimension' + CAN2='Second Canonical dimension'; + CAN1=0; CAN2=0; + output; +Data Fig3; + Set Can; + %dclanno; + Length TEXT $30; + %system(2,2,4); + %ellipse(CAN1,CAN2, 5, 2.5, 103,color=black,line=1,width=2); /* H* */ + %ellipse(CAN1,CAN2, 4.6, 1.56, 113,color=red ,line=2,width=1); /* HE-1 */ + %ellipse(CAN1,CAN2, 1, 1.6, 0,color=black,line=1,width=2); /* E* */ + %ellipse(CAN1,CAN2, 1, 1, 0,color=red ,line=2,width=1); /* EE-1 = I */ +%label(-2.1,5.0,'H* ',BLACK,0,0,&ht,,6); +%label(-2.0,4.0,'HE-1 ',RED ,0,0,&ht,,6); +%label( 0.0,2.0,'E* ',BLACK,0,0,&ht,,6); +%label( .0, .8,'I ',RED ,0,0,&ht,,6); +%label( 5.9,5.8,'The E matrix is orthogonalized',BLACK,0,0,&ht,,4); +%label( 5.9,5.4,'by its principal components. ',BLACK,0,0,&ht,,4); +%label( 5.9,4.8,'The same transformation is ',BLACK,0,0,&ht,,4); +%label( 5.9,4.4,'applied to the H matrix. ',BLACK,0,0,&ht,,4); +%label( 5.8,-5.8,'(c)',BLACK, 0,0,&ht,,A) + +Title '(c) H Matrix standardized by E matrix, giving HE' + m=(-.1,+.8) H=1.8 '-1'; +proc gplot data=can; + plot can2 * can1/ anno=fig3 frame + vaxis=axis1 haxis=axis2 vminor=1 hminor=1; + axis1 order=(-6 to 6 by 2) + label=(a=90 r=0); + axis2 order=(-6 to 6 by 2) + ; + run; quit; +%mend; +%macro fig4; +Data Fig4; + Set Can; + %dclanno; + Length TEXT $30; + Drop rad a XP YP; + %system(2,2,4); + %ellipse(CAN1,CAN2, 4.6, 1.56, 113,color=RED ,line=2,width=2); /* HE-1 */ + %ellipse(CAN1,CAN2, 1, 1, 0,color=RED ,line=2,width=2); /* EE-1 = I */ + rad = arcos(-1)/180; + Do a = 0 to 270 by 90; + ang =113+a; + If a=0 | a=180 then len=4.6; + else len=1.56; + XP = CAN1 + len*cos(rad*ang); + YP = CAN2 + len*sin(rad*ang); + %line(CAN1,CAN2, XP,YP,black,1,1); + End; +%label(-2.5,4.0,'HE-1 ',RED ,0,0,&ht,,4); +%label( .0, .8,'I ',RED ,0,0,&ht,,6); +%label( -0.8,2.5,'l',BLACK, 0,0,1.7,CGREEK,0); + X=.;Y=.; TEXT='1'; SIZE=.9; output; +%label( 1.1,.02, 'l',BLACK, 0,0,1.7,CGREEK,0); + X=.;Y=.; TEXT='2'; SIZE=.9; output; +%label( 5.9,5.8,'The size of HE-1 is now shown ',BLACK,0,0,&ht,,4); +%label( 5.9,5.4,'directly by the size of its ',BLACK,0,0,&ht,,4); +%label( 5.9,5.0,'latent roots. ',BLACK,0,0,&ht,,4); +%label( 5.8,-5.8,'(d)',BLACK, 0,0,&ht,,A) + +Title '(d) Principal axes of HE' m=(-.1,+.8) H=1.8 '-1'; +proc gplot data=can; + plot can2 * can1/ anno=fig4 frame + vaxis=axis1 haxis=axis2 vminor=1 hminor=1; + axis1 order=(-6 to 6 by 2) + label=(a=90 r=0); + axis2 order=(-6 to 6 by 2) + ; + run; quit; +%mend; +%gdispla(OFF); +%fig1; +%fig2; +%fig3; +%fig4; + +%gdispla(ON); + +%panels(rows=2, cols=2, order=down); + +/* +%panels(rows=1, cols=2); +%gskip; +%panels(rows=1, cols=2, replay=1:3 2:4); +*/ diff --git a/R/cars-colldiag.R b/R/cars-colldiag.R index 37c05696..2da63558 100644 --- a/R/cars-colldiag.R +++ b/R/cars-colldiag.R @@ -25,6 +25,11 @@ cars.mod <- lm (mpg ~ cylinder + engine + horse + weight + accel + year, data=cars) vif(cars.mod) +cars.mod2 <- lm (mpg ~ cylinder + engine + horse + weight + accel + year + origin, + data=cars) +vif(cars.mod2) + + # SAS: / collin option #colldiag(cars.mod) @@ -121,3 +126,20 @@ ggp + geom_text_repel(data = scores[dsq > qchisq(0.95, df = 6),], size = 5) +library(ggbiplot) + +cars.pca <- reflect(cars.pca.save, columns = 5:6) +# why no ellipse? +ggbiplot(cars.pca, + choices = 6:5, + obs.scale = 1, var.scale = 1, + point.size = 1, + var.factor =10, + varname.size = 4, + varname.color = "blue", + ellipse = TRUE, + ellipse.prob = 0.5, ellipse.alpha = 0.1, + axis.title = "Dim") + + theme_bw(base_size = 14) + + diff --git a/R/collin-data-beta.R b/R/collin-data-beta.R index 6b9859e9..5a2527a2 100644 --- a/R/collin-data-beta.R +++ b/R/collin-data-beta.R @@ -57,8 +57,8 @@ do_plots <- function(XY, mod, r) { confidenceEllipse(mod, col = "red", fill = TRUE, fill.alpha = 0.1, - xlab = "x1 coefficient", - ylab = "x2 coefficient", + xlab = expression(paste("x1 coefficient, ", beta[1])), + ylab = expression(paste("x2 coefficient, ", beta[2])), xlim = c(-5, 10), ylim = c(-5, 10), asp = 1) diff --git a/bib/pkgs.txt b/bib/pkgs.txt index cc14beb4..4f2564d7 100644 --- a/bib/pkgs.txt +++ b/bib/pkgs.txt @@ -143,3 +143,13 @@ ggplot2 heplots knitr tidyr +broom +car +carData +dplyr +ggplot2 +heplots +knitr +matlib +patchwork +tidyr diff --git a/images/arcmanov1.png b/images/arcmanov1.png new file mode 100644 index 0000000000000000000000000000000000000000..391892da143e4241430a96807216eafae9f8a15c GIT binary patch literal 26900 zcmcG#Wn5KX@He^vK^jp?S}6hH&`2XlNh4i{5b2OUaHyYz(jgo=q+97uX{5W4ba&U? z`23%H-`scixi9!E_S$Q$nKf%>*35VI`KF>IgNIFy4FCY1oa{R_06-rD05oDuRIul; z8tDf%kIdgHz6F3^5jZ!-=m3D_Whp7CA}1+HtKw+?$NFHU7j>?;a7oB|#@rc*ObQ2bnezk&3EYq?EW0 zsTLvecZ!KZ$I`QclC3+9`F*WYsFqvvq{ck(q^M}uX2+NX!1`t>Y%fEhX{5q|BZOz< zi$;`=j%hqM_R)>zF_bnkr7?-(B1kk8#o~|XkS>LfxCjGsOT=d{eH~U}Sv8UA(|I3+ zi6qR|_cI`@K(*t7-~q3cTax=LR(SVFIKD7VefNDl`DTmac+$O(-}_!Gw6vNH;J^nc znYZj4R_*)4@Nr0A_dNM+edjsPn_s${wA#b>sXRryO72ijIF1Rz%*7}w5o2clF|Y7i zY~a9$mF*>~?VlM&2J>T+A3sL53-ozO^bo~eWg%Z9%mtVnEAM9;D#^Vv_iOL>`g+k? zDr6%OnTuX8hyXz}z$nh=yhjLCJPIw{JRxbI?_y1|*u_Hp$yh4NZ$DfWpx{5f7L3jo zTx}H3B~)Qw2qtP0K>F^>qZYITTGE|I0qKu+&>xHQ=Anh9Kd6Yxr!r{fIqZt(HoiUcxd2y7^CaEG?g$D_!%_x{B1_bWEmnj&4mlxbylz}E-% z{R)>bF6ig5pkMaeGC63x@gOgO7cUjw;SWo^`O5a!?-5-TCc$(#5nZe-+DPCDEm}94 ze?&iBnhF^aRfmLSM!$;ED={L{*#Bn2S)?>FmC^-OZ}T4K$&I0Iu_lJbNm!>V4i9XI zI}-^$Gx%Eajz2?ah<_+^!}8=YE=K$BJfl!frFdF}*PmE(h@8kq!#6u{O>;Oxv|npd z7W(mac^I|-w49RId9fq6qp^eON`6f!^SkHo)SJip9PEUMh{E5MOC?Lv4%{_n!o;Qi z{2iW4?T6DBTyVOk$36sNVNgcQXX==1kDwB-jhQlnhoo!Xy1#dqp?|vijqx4C1eY^H zTS{25K`NQv`{l6g7QLPtCUZ(fkaO~wyoZ8^3}Z}ekLHSzBRPzqHUXY;lYFE4l|74X zL#(Pzdi+$jl%{akB0Yh=gec@=q;7XB7d7D*dlH=^US z;$yY_wSQ^XPvuRSe~znZbWgqzJ!C&LfBFi`1q<>`;Tzpke)=_r=h2^{^?OWu%A=LD zjF?@R?emM(y0S2`NwUqdsIp_S3FM%NRD^M_cF*r#^=|#BgqZ9;-R{nA&2Gb3ua2k2 zI^}W_@{!9Lz9HU>S-#8aJSJn{RkOpsp*WYlX|A#8`LM=lgL)rMPs^+i(r`9ix z!*5Rwzhx2?JF8fVTgvH335dh09T7)G!STW6@5A30_I}VPE6BDAw)(Oq$s0;8!=|+R zbH>X>dtOCEw^8~w6{E;Fi!k%EwyE~BV#^;1dDH0vB~l}06TL%Lh76aPDZ^mjS+>_gv+#TNwR4WWI@-Yf2uqVK|c)1k0 zMXaxa65@1{bwZ|h1Y=zYx5+0J6Z(>Z5+;+5*|bSANMzV9;vD0QPbbdUmG4$c?ojRq?*{Hzo_U{*pAAy2Q2n6Pr6QrkqpB7> z6U<1_PdO6^6OtElN=Zr~)ekkOsy(TX*7wpcG=SFR*B{P#%o)3pp5N{%PqG#(>o2aN z2q%l2c?ut^A1WL+Yy=!0ofI5;Y!O;Oo=@+L{ylD;?5ys0Yf~PF%$E+P{iXT4 zxYJBzM5;pFLNJCkN8-*e`|=fIo)8(u4$T}0!tk&1FY$-@^ZK*deO^$jH}kXZ6Y2J%vLiMm6eTsLR;A{AW=X9> zEkI#LT`6Y8KEqYV(!MN~W09ovqH}Cp05W+^@N;xI{l)nWy$W9eCeQ z-mf{x-kaL@K&Y-I8m~%{{B%-!C%KAku0O`+3>2 z`~FLB-538an~yIw>@?XW|j^h*;EDMVdfhjlJ- zLC}qBW##sRYi`|a#mv@h!Y)H@>*mR5>41OgbD?dIs(s3J{aIM{aV6DB_+O#77aLPxnvYB!_{(qm9i``WQ~2OIgmHFmsEK#SBf#&f+?${=ZLc8>qE z$iY>+cj;XQMG3#%ymJlAtHon!i|tqsVVAVfl=m0AF`-O-&tS^K z${}g0X&RykkA;K6nbKW2-QB8OeNQW?D1!M@_qYOL~B4xz-Dzrb)ipO(=#8po5t4bCaJ-Wb(*K5 zuOY-fMECB0pRJF!tosgf_a(4S#TG!UpdL+i-d=TyKkE|7^9S#?S%IjXM^HbgmewLO znr|_%BtN~i<#@Zbg}?wJu>gl{y`7Jp?v&=blId&d|^ z6(v0b$+S$I*|Lq#s<7SV0*Ev7&$`KS0OXEB#1A|5nod` zE+yU%HdA^vq5>$f{CsP$iFm)CF4C}zbH_K8FDsNTBv1vKo_P(T0ytrK8_Kl#L?$CN z?cuVF2gEQ-kF)H3v!_Wf%9KCFMohO}@I!VGSnw@ZxxHJyPm<@9d)&^@2Hb0 zWY&5*oYp$uHhn&8$FV257WgUfiGefdcHNZSH%-mUJjd2X|Fk^P96GjbB9_0w-`uF$ zI$sFyo6Its@e6AE@{JF@zQVyhH&tt~gIsD*L?9!m&zdeO_8=g)%@%RsHZfS*U9l*7 z-Ih$eY~yjgwd-^R+j*@!ykny3cz#vUnIKiCvi5BU~z?diEhJ+Wg?m_+0jJcX9NVO=3MqWA_NacceTnfvNzP=)lHR`$L# z%lVC-ILM$5lC-h*Pin);GURVt&;rR16KY_v~+Qa%4-$NiZh%&-*Q5YoaxAoJ8k zz;K5RJv?LdT|AaskCg4NsvJTPTpR$;aHU@wYeBZvo}R8-u1{kp3>zjeQ1&-H#CN`U6f`5`)3Q9 zSl$nL^F4;_(uq6fvjk8E^c)gjt^5(<@BLuQrcze+p^M=#D z?(4c~ToHZt0?}!QeLn>QO#4DXDCs7WM2IW@ebc>>yN2Ix;Jsj(VToVk!AC4v_NHH- zH|qr#NwqHO1uwNWvOVZO`K7Jq=aRhAnXYvjc=1H;#^!j=j&4^hNkznmO1(Q7@^Mn% zz~xMHDQM5-utqn!MW!vYsXR4jCFo=w8?DrS&WGMnu(}&>p7p@|T;XU7Uv%&b-}c8sjr z+%vLS+3Ps9kD3mxOK_XK@qafqRxzdcry*8GJXpjO8RSz}Uf3r{fC!EuAKCrYK zXBPee>q*>xz2K8!Up%CwPw_2JRiPKmlRsz zd`}aEy~>&4YUlPgKdR}mG(0;C*~nUadH4W}#8e11#uF&$>dE2>O#ROykHGskGAb4Fa1TGnf<81{Pn#Z^ zkMwX9u<4e)ctCI>qYgmC2B4)w;((2~bQ>wy_=fubf1G%`cO6%0Kd+q1*F*mAS6Mv93&8VWTJ>J>R$wca;PRGUhK0-TOUqW~<>H?k+U70VH4xdwF&^Mm z34Ad2aUDBMk>tm|_zw!Fk_k&4j7HoJPAc!CIhERYyVQK_@8!S0uWt#f+zmN&!Vixp zPOWXUj@lL$#_tm&n1r2P`1^Qo%3@5oo=y9kudhezlm^H5z3I8E^P=~=`?qNR?d|Mc znjej!3*T`6ZYyFM9v6Cw+fae~a38sCt<*au)1zC)*u1}r=?lVwFuT1H zYFT`(5*$c@bZPVIaKE(mnpyZAMz%UDyylA~PPw<%;V!W#vzpPbatUlGv5i@V@66=Q}LziFs9eEjHqf&7>w3A5+`X&u-m#;h6tM{t>7GVG+QW|MHZPnI92Dh7ONTO-I-tKS^hb^(oXc6t+6}EA{v4 zk^8*!QEDY&5Y;WEf~V_G09c}11D1Xu1F6_}&1x3Yjb8(RztjK(2cSEHV+keiu{xR* zx0r~nCyVVwJS(5i4v$E)yS=o66u-DRJ{2>?h*>0apuW4%fenPO4|+w+9S)v)5fmNl zEF;G|2g%${Lvv@?fiY&`K~}&9ck-B_GY6{BPgHB})Fhf80V(Eil+@~b-h6Nz*L4@! z`P#{{o-+yaeyhW)`Reetwf3_=!Q^pyq+Asa5G%ATz55Ln@N7SY>oj{pa;$JquBVjo zoTkwJ)=wETg`Aq0&$F6TdDfiziVM(@HmUz43z}EvW1X|VGCV4~b2q~CyCYvczwPYe zmzpm++iLm^UZW}nfA8~{ z27eq9`^JIprR7e-@EiA!D53)ABHjT(G>gKp5V?u7q$W4p#RYaSen3d$Lf4x4#Kh!{ zz0a5x7n*s7IM;gu)nd6}t$)S5g^=%wBV(t}!Hn+L=>^VCC$Ce!6+%-RJSr|8c_I<tIiNP`Quc)2Y4V*(0V^T$;%C z|B4X`YDf`%5oz4!OF5<{b&*`mP@9KQHP){(2(_Q?JAc7DEr`b zkW^hY)LgH&;klr_l@=89e?3=t0PUM_Jd-$3LV!j-5ep-DD zKXHsJcd?60P|k@EXfRk@7d4-4G=S-|t!KyX;JDVTWZQPw2zmz$_C2fJ*W%~7-wlFS zwNa+-^sTqTaPQw*lA)jp-b~!quB+-ZWqBrA6 zITXot_oQ4ap*<+{M z`%&z4^UhtYh3eeNMeVWO6Uz$noe!Gz9{(u3CaZM_=ccF~2olqOXGpIV@>r$QK?F+D zV_n&ey9&E=G)%~xdlw#0cV2l1wN*v z3p3mC@wBn{12z)j7(;oV+erhT?g`)mVdTfs&i;+m#xy8_FX6_tNiTjQkw$<@?VlBs zATfJ+g|N9r!~L+I5!IGeC}B^VIQ#|PfIf)Fyk9AWZs!#98j~ZPkZI0L!JIq7Vh&JQ zn66kIqTY^*&#GLaaZXZMYGehZGL8hCw9Fid}F-AwH=Zb81mdm@}f9OwE><+aDejx?$hX+ZHbO<9#%+^vUmnfuG%gUJM6UCEi#D z03dHDpVr?TEOdcB6c$B|9ZG1pSk1J!?EvF@)+s_LkeWBcT~8VLhrlQ9rhIn$g2B|4 z-uqbx`zK!!5$K!9Ox>$#<2Q~8(y`N(x0-xnn-`d96&61A7c0ws$^)JHM6X69%}xE= z`-B&_XVg7E(G5z?(B$YQ?RB=)N1I*#n&l4AKIKZuj30M}!@f^fmWNL#+0_|P8>dsm z`g|aMWwfGS3W$$# zh4qLe%S^Bd0?wamu-p(`_(8i#@p8=*1nhj=?q)NZ?g{=3V#!bLtP;-U7p&i6OgwRS zCA!pR-u3e;Tgm?7ByM}gZ6$K^38$apy+*V`ec>z)@C9E; z&}QTJ{lS5jx%|SyM7xA21XiH9xS``!>JR0QMzqtU5)17TEW7P)=`?MbxBiXL>o)-u z=F;xN0d81S<<)JdyBx03&)NJd zqNfiF0EUv0Yj6P~_u$IZUuMV0-3)0q)@cbH=civGuo*5I_UAB%RJd@{tbBjCaH!OP zxBI@_W*E4_3MTyN=)ma9ki6~%;Zxl9p$x*Vrq zAP*BSXXSo!+7f6Xux>Y*nt(CEb_b?IH!+ z+ojM(1B4T>ZSSXB-q9oZS35{W4~~PbV+%AKGB=;zGX|xx*1E;f;Y7M}`m*C)Rb@(= zpGPiQ9j&(^Sk}^J1?M()D;WsOW4)xP?`&2A+q%xIVm5k^*zmIj>}4~NKCuYcjb^WY z7+^3nSQ7g&!2G`du(dEvq)?pB$9+8JmIbo-`HjBD$7{B86QJ>F6I`^ll?ZA6A?RKV zS>-*UcmCdcJgj(5=uN(;_u5wYZ=PY)?YYu}R4#LYwfG=NH`nT;jh5;1VD6yBAIHL_ zLGY6M+RL#kEl!HY+tyQ)?T8>T&#DaAs@n55Gx_=f_z_jm>3dY`@Ymuq9f07csbQMq z;XeI1XbPJTNr3R8rcuJ5z>+xOk-G4Pir99B^^nrUaN#P1CdI1-?`P&ARCb*3vgmH- zIR0${%HwVvE;j*`I({Z9bn};tN zyPYWqu&bcQIw*Uw+>W+@!N~Edn0fViR3Eee)wHV8n-P?JR8oyWmAoFz@#9|ci_2Dm zn~$O%UrAv(sj{ANJ|!%7Ga&*x(WJg<(Tf3$_EY43;@=4F20q^=JTUCLxlPUb9O=6K@k+Ts7&sxE5WVS%46lw^DROnPvmWd`tIfM>|%n4|et*Cc2WA^kh(A-Pk)N%&2R0;seH%OyNRGKXU@{KsgL+N6^p~>iYdcp#u1Sd)buQN| ze%mwTB+K>U4h6T{3nTl?KADxa-b~J-Tc);S9{eb;_*-~nL2J|~gRrYbYggmL0h8L@ z|A@X*Q@uVYf<(9ft!Z49Z@IP@o-~Avjt6*u)gFC zK`C_!cEp12&1c!NyW2#9`mQ}6GM$0$N|cj=ld-=?HUMq?5f)11>VgR%W*mQyFQgw}~1>tfU0nCg?E<2|Ul_ERR0#5t7 z4BMqhTusMUS5C~=qC#(w)${9&4=9{(UB!rH$u~!gvLusL{vAvZ#b6ofP1aHT&Ycs9 z?)q*J#0q>I=nk z0~g*ZgYh*r+u}5buJzd4lQMhPZ=10bGyXhi;(x1wW3`XF>Jl&zX-uRtorc_0fwB zm~Yo|?rcZX3!PqS45;f(U&HWzm+}b@y%zW<4DZ>w)cJZndCv$$rY!n`eB(amhgyYP z(Vn;3nvQ%JyEc+$1D?o2Rzu{A`-$kB;mg9 z({e1En@>Zf$TgA6rRyM0CQh=LifR5<@}gQq?z$wYWBP&l=}iRPnR@{OtkAUD9^P2y5{=9T(cAa9h&~ajvO7cyE_<=TVbNdYde~ zdMV17=E{G^bm_Xau4$NdzFvH;{14gk7V?xD&Q>ilvX=VXsLn{9aFB>FXwRdTJfU*EudWRU2%ff@m$O?Oz0{XG9l;`TT>z<$Y|JE`c{a|sjdcWuX4=Vx0 zc9_29AvIbF))NKh0{pKvV8TS3m#TvLj+-xzZr5`pW%eY$Z;jpee`dKUgvoCrJiQ?K z#svJp6HVBMjHACOSw~kzc|&>SIsg6RxAW3*)ls3~Muk1r7fZH@mSw^c~_FP|H`{`a>4Gbn=(z_Of~E?3a41po}( zc+UZ$zJd4F@c&D_o?AY@39k&Yl1BtG~oMT14H?J5y+fF`#zZN9^az@ zY!C470x>p-7Bt=B?Y^?*qa@xd>PblJAkb`Y04iLa?i zP=Hd9@W#~9`0cU0kav%u;;3`wO7XtqB7Pe}`6S;_8pRhIgx807ytuEIZ68ck1*bOa51lhbk?KiT@bX|QkF-Zb!aiWj|OjfzpFx;&`5N(!T`KH#^x zQ!C_%yXS@){W2!+N{neRHqVZ*K_?mLOqTguz2%yqpbEveX%6#)9mI7oI~L+}7$~$H z8GC^4Yu;2@b43BX+Aw!jG14Eg;4TBNz^?`1~2!lNT1N3yd!%ENS zs^PNeaw2IQTp&1a1>Xl6+PWNHeq07CB&2)N1USD?fstCAQ08!$@>c=0)fLry{A$th zsG*cy>b(8tPA@0+t8;LXIO59#Ga{wxdB-+)20%BHr~|Wub>sBAkJA{G`8;i+M=>Bd z#eG*7tT(fDs@tFtekffE>a*yMvn{0ZJqBTD2jwXxJB@U3hLFa&Vk1LLK)lQFt@C`{nhaG%JhH`K)al7&(@Y> zaeMPbZcC8{JRprN-Yo5P2r$H(&?oYetPZRN<_*#wzB8|V3fvHXHTeBF>K%tO7)_`o z?8T6#K!FX!6hm>l7V0-63$Rj{&X}|&RazX(;4^%Q?#te-e}vT#HS;U`MH{KFuB9>{ ze)^l;EICZVG14;NVYECV^0V$!!De8HHldI8j4-a*oGltAaP!UO}IjI3b+v3sN2O4sE5*nI#tSQ}A^lU_*7L=h)6 zLG0NI)^6NO*n$qDyZ@L3aF5tD*g_OI1%vE>qEd;>p$5LcbhVV2(Wd{&cZstPmUdTG z=0^aPxxw=#m2Y!?-D_EWPRP7P}-524W0v@KGxox7nZFqYu+~*<> zsiyb_U3?UOo4I=8%d=9I7od4iyJ%f%Q80AMqp48*k{|7@W^ndK8ZgeGZ+wX6`=djE z*jFCmN#gMWY6gnx#PvXJ{Nho59vsdM%vJh?RZIoQmOL{A6>=le_$_cEvm?=vX8R(H z;lV}^sM$UNFLz845r(*Pf6I_;puUaeY!3SDf3wv74Jo(Z7xkEpdaMZO8+(x2eXW&n z_RgROtAerx^}09LR=^+y$}8nvG-pXrLxdOXr~;F655p$QgCzKANQ%Wy&t9OWf68&& zoDDG_4fxFpN_}IZdbAfgNH_Tt4nL0CAI!BJfuqn0P#M0GHVrfZ2gv3&amrku{SG6M zj*f+k)-?hLN%)ZEZnsZ9CDT$;Fz}o|s(c9Zu7?q(YPcofNFSs7(mT5&iOIRYXMjPT z)W#WVi$D)fevM-u=&EQ_V0XtfDB%RaJ=d6pcC14EcFHIbEl5+$mK+F&%ls_ydFlFG zafW;EDnl2HF2lW&AV@7m8Q#yhZ%~)M;zQusg8N#N%O07arvjs>V+2m4((p+V#6)-N6bJ0D?VW7@~yF}_T}FA<=G zJLi5I9p*$?qAulmNh`c?x&GMR-g)|mn^KHM#(@+Qd%Dl@+~Jd?Yg@uFGLH;SGy*xp zub-L!`;0aG1XO;&duk5DT@V}b)u6_zR$>GfepktMD2R>;zAE{=hiQ6c`@ZAbB@VOkRsCiF z@PTJQiw>%`V8alj7W2cxBRW=lt=C0XtpIM%TpeJJD}+kRyldlo_|*Vtd$1(aiUG)? zbX5*ti#t685Cct4{AJ<{u*}(xBaRC)UA9?s+{bg;;0CYvYkg-raf#rfkMNNK;o?#ykJ z9^;KF9*wk8qc}uOp)c`)bdEMY9SnQtMg7lM&v?(D4|jGDk2Y&tlcc=^A4@A}8I4&tXWhfYoFcjGb2dWjjkNRxk2MJ%x z6etq`P>d`pOHyMSYDkk78wjo*!T4wOQ+(QaQ*6l5-3O?BVdkDY>SLHy0btc@m zN2muRpvZv}W$?e35C1tU9ieibald@a@-TQKVFkV9$Xvn^^gnqE)x!;TkA92A4Lbv5 z@l)vOPeK2|60jT#Yl>_JIOstNBA~Y9MHqoG5GV!>Flk9gzm`58{B&acUpYg00jlIHv(Z_zsQQ?`PkPo@&A^Go5vmHe2-LR= zmH&X;2WlR)5xbmukMoauEa(R(JKPZ35CI##gQDubkS0C;dv773+eD?%qb9ISvDdYIqJ4RaKj6NnE@*GHb?lTIy;5KQ&F?D5RRAv+iYcm}?B z_-bN8ed&;$bNq^5YgM6f&#j`PZt*qio7a3$EaIx)qw)OyNw)-z`tqfu>Hq`QGvIha zKGV87HpWej?F$7ZV+#uG0xl}Gpk1-~aX8KOTq%Ysof#C$8Ya}poni66h5@J`1#-mh z2vz(8IG*ikQ%lSHPIj+Z8o90~zPMl)Sf^-lU-+Y6M+48I!WE|mC%a-njW~i@gC=Gr z4@-Rt^ly?nP+rL&p$_pNaYD1AhdB+ZNbBdAfh}4XibaZ+Vmk+4Pslm8*zkEbGhp=+ zhO*&%fFAg(V*EEdffq$9Sou4OX%K%NlS7q&1x<$reF^26p;qHWQwd%{?>!EZW^5zf z80Ohbi3(W85`W-66|~|jHJIQ5n_=dkuG=gDATkI%Pi~JUaTyj^!mx5I)4L#$_wNt1 zB2pnk0qt)j_1K<0Ch@hQhp41ZyE2Pkpze+$8Cbmxfmc#TsQX7UGNNsy7CNlI`lO!) z+(b7357g65!9GpBTJg+z!wD9#HlRY`sO->*?5pI-OY_HePnyKPknbKGjjO0mcGYH> zpbe4t^y#kbP%L&}h-n6ND2H+>evr)V^oM+Z0r#7zh;MjpR??hjZ5vDC%O__@!oTMy z#nGWJBE*)OSoL1s$saQGy(*C03l={V$yoBrH(b_`MA?I;!aZ#(XaLaa zZuiJPL&xw{qlbjhFN@+*_aHA-T@07NKsU^S6Q*OG*D{(_jStjx(pUCv3Q_5$kzIT!=?~hC=q`UY z>N*?>el50QR$O)M;9C^lj)<@4jeYzv%4eq5GLibOHMJ^EP}cChfKcGz-~gS{^Twg* z42f;|fl~~Q#4iU7Yj(EH{hOu^w1)Xu_i1y;F40`8z1eglzxFPUwH<%b8dq--_bee@ zF7*L3&6Yj_Zt~m@Pgj28LGgncy?ko!v2plKe66EhDK0OW5}H*1%F`e}AGc}@&I-B)Mj_`M3lozl6xA-Jn(HoLoPm^gCV zAIQ9hkt7e(usts@X(0q=4;XB{(i>4#b>$5uaK?2nKF?i}yG?A!Cd4U-DBI;jvwK2< z7yvWy`?QAQ{FQ9U(J{s52)jYThV>oI+8=rrTbhg<6u>2{k{u9x)LeYjCr#vN-i~}cvQ9{duSl*X|<<)W^gAlS@F?bo}#A@ z%%=2XV`O$oZ3UNin0epeK|}xh68e6X^tyQwmI}F5wcppPlQRUvdXo1{5E}EyRFGU; zTr@&U;7vovjzPm;_cnuwZ?u_ZmmM);E4F{Ke!OQHcPp~DnL2iB)uV^-I;<3~j&YQM zUQ}$BbrEX2*IVI8`#woTCa2l!9C1G(=2QN$npbyc-PzL^(B*2#9v3{^4y$ds>CouQ z_gkp~zYr8${&{pzCFP#=2Q>E=;21L}fl814nY*N1J&~i+g!S5eh#Uet0EA<%HV>kr z`bgO=ko#t->*XzZX)?d5b|i%74VNHn&&#Zibah#7%P!^5-7a3UT-G}@9>&>TIBqKc zh;F8_=o~4s{A+Ix<>po0Y1Nl)rH6#Rl@HW%<{Iz2%aG6b!`KHU7roGMw4%)zD$Maa z(S>E0F81{KDhq>-Y}z@x0p6&7N8eR^`31_5i=hTyc|zA2(v6x4jeDd(wsUL)yfUHL z+3}!oY&2iz&Zv-u_)CqzI?hmTlTgEJL-L6`@_9or!hs)s{cD=NkLUQA)Sv|^tP_=n z+T;(3gi0Pa*xY!iTDR*$?xCtoKZ<(-0dnKEHA?pR6`{2N0igpBC*ZO^Nr=BH<5qel zt|YQ(-ZU)lERf2%ADa8qbugS$rU7(7MR~yhtwhj=#vKig_tmqev*_lJ**THyD-{*} zTSdtaxxq=7C3^R#uPdvC`;FEC6S~-SjR#;~)?g>Y0R zzV_hQ56cNKn}+B4r9wCY%dhnC#yFWsX{mbT8x1^!=$!nNu%NUN1uO&YZO(I!*fck9 zT^p9?AGyzd?di)6M$nupYoz+wLDa#lIHB>HO-y#jk$J0tg;d_!lH}Yi7{$;m+Ti_% zSV=d~XXo#klI|`^ilmtM3X(aTq)iAl|K_J&n5id;V0i*&sa~6+PM5zclLuU3$J5fY z;W)SY09QhQ8rxQnW zIMycqC|8pfyoCD+V-o}~WMR##b6hS&Hj0|q29RExj;J6}W`6CR{T6@$X1$1oqHQX07^Y#eq0ivu~`R^vT@E8K|CgkgYZxKd(@^vNe51AAg zz-yy7%(chY&&v-VB37;O0u1-wMk8>_)*!qe+0B!hPVRmOCQ^#?wUL6zS3P~1-Co%X zZ7x?!u(soyxEkNtG%#hxc6a+N8o?iLg}??6G`HQH&Kj|`9L^e{)%o0(Jj6xmx=etJ zIPOV0q^@x;YD;`Pt=7gy*8}-N)@2V>CHMgS==yo<&5cdhLdrt+-HjQuu`N0p7#*(i zD!TcGSt{}TqH7%54G;talkXSy_n)`0(5%h`Q)Ql~+eIUIN<@1s1YLC=9U-b*A9aHQ zC?&5IE#Zd2ekI8{?KfihhRMh5pApu--r85fX?TBuN~C4JQ>Izd`jJJ{w|es|TKc9h zVu8v{p*d9Ghwj7gvs#P_-WyH?OcR+VVw=6xpnBr_Le5Z47k^2I*F(E~!0C8=(K}aI zA4St2BK~$918>oU=#|2S2tE~$2nua_-_&ZMw%&hEb&V?A?cu&^iSE`%pITg>B8ewp zlt5s{th`51(65=t99V3ZA+R}LDkM%MRByMqyjn;mQ09sZtwQohO;(T1&mA%_N>BlA z=%c1n5Td;8T~p?ZS8x{_J)1olu-wxffaS9BgdY0qwIOA#~TPc$h1pZ!y{Lw zF7Z$4W^JUjNiFsMiB9=3^4mf<96*n}Ya=5hWWY7|GsvD8Nck;4mEi#8o~lnh7I1zt zze!itIeCe$eOO>gTM43E0D11Ly!5E^%u@^gZx?%GyGPaa%i@Ay&@NXQ-S9hf0wY}} z8WB2>h_Wb38%AwAEO!b@5ep=ovKCDW(8fA&r_X40nMhB>3t+Iw?H8U6&?2EmJ_j&^t zD`tmG|HU(2t_DO?L<%!uX8(gzdr0C3B37dO~>h6svw)mI1 zEQYv(MQe@SAOo`R%Vkl`D@&0h*hrkl3!S&0t_<-%a%zYS8n>w60%+mU)^5KKl|^>t z_vs5NWG^JHt?rI&&4dRS*!I-_kYWK+vW8E)zc)0DRA7q6ezsr4_L<~)cOlV-diqxn z`g>bIg+#=~Psf$5LkTQmoLde3@u*_`_B<3rJYr5F_j~=Ie9u`lPNvyii6!{E3BDAs zbxlX>#t0*iUZ~oeLf+Ew08#x2lyLx#`=*y=^ziklp@$27ndWqP`CyGayHA^j(*J1L z2K?$X`A}tP_2$2PoBtZi3(nI(P#=|p1?`y{i%MEtQI55pAnGmG&R4!l4Kr`3VU5LS zOf|EfzP|?!*eqnIKe=D`@d$yc(yCiVXa5+!)(ty^CShiI1h7%oat%v2EpVzWnW1+<=)oUuu8I zylf0l!j^T*fxnooE)zV%CM zfA#j>QB5`ByXX!OkSeHrbVX?jLXav&z(z4NmGaSoNbd;JTTqJ9d!%ClDFQ0J1q@1+ zF1Unt25pI9Ff=rSMq^NLsXs=RJG`Qv=PG9U!^uGi=|f{b*^xHmXw-o@f=x){C@CaY)i z?QxS<`&Ta7>s~PSc*Kd7uwr7wTI)7t!<8d*rD~j#;ry*>b5(HCMPws4F0tLsN?+V@ z-1=gsI| z1H6g#Ps?~y{>Kvuli$T6BI7K~ORHNh*c|J5u4vR-z5IQajMDLKo9h7$y0^ca2wArPcQ93g}>qPAl7R>c;Z|@wD zmFRyXM@y`H=@2aQYs^^PP#1)Bae2S)3gik2U-lgYV^VuWIYx+gyWCN1KtL6(0y|Dq9;Bhw)+%IUJm5v#OL)6KIb-hQV zC0Lu<$Z)r<2OTQMk>mFn^To+yd~c)Xts%cr;F#G!VtQN4cXY_NHrH1IW%qB!1+Lz2 z$UK{Rb_s3go ze8VRpO;e>jBW?On1mQ~nC3aVz{fwH}oM0DjvomIUb;SRqoag3bU<^CtRkyk$k&1K9 z4v-}P>W)uN^|W{;cQBcLO?>s$PgGG_Q0tYZ`aj3@+1F;!aez^4fTiWG|PUQ_rEcag=>WpBhZI!&K}8cF7E7f5msL@L4eG-3Qr~ zZN%@jP=r-qKiiC7{KEzJoXYZ1>-ZM>W^)Uxmp~w3 z=~NfMGv9FK-il`^>KP&B-4%%4>E}M%x{iZTbCkR0{_wr9 zv8%04yW>xT?SzKrJ9~E=0t1VKM53wmf2?SA?2TG) zwos(?E0sIuQA$E%4DLCp4VzJfIf^gzy79oj>S|l^Zq#mmiZHi2)U$cG(Db?e$Hj5i z-`h2G=-;iYJ0#f=*sd-aY1|*3uYc6~axjd2;EiSMFw+EALQY*{$7t1q zOP*f{c^9ToPjvR~Jd$qc6Uyk4ZWE!5HP%642?&Ui>(F5B5!;*OHeuZfFE`ciVsH)A zY&@{n3iy5Pt8o(zM-PuF`lNwhc6hg#R`u4&_-Iv?*+?yq9yXyjZy1q7-41-bWDS(d zMD-ci6aG8}rhwupm*vU!;e=HL5AfMUSM%Jao-5u6RO71Tff(JZcV39ak&m}ELx%g4 zs$I&j_=W)bzLi1lZ{NSp!|+euYDs5u)mpjrKQ2{$$D4c!UQalLw`=a*xwu?{=s=z$ zv7vAH=pCnbC%9Hr-HJ9->M!BL%t+QXs=SdiS9g%S+qS=O_gZaJ*NsbSyp(UTv3+JG6AV`)a+f zOa~$Pb%~*Mh2LCe5odossG97w70cM}GJhcxjJu+D<~(+zs-hdBBRRDuk@Qf4)c&lsZ9OfO8!g-Y!#gji6oX~ z4)H!3Kfc!W0LS{*g4?5hJ*aVP(+^f;)yy)<8^ED+BwKpm&22ymLj@J|2fa_>ri0YK zJDYfQ#BBtTu%D4Jy$)p&`!qi=U~2e>B3>3M`;v=ff{pz1h$2u$FFNcFT?MyHMnNz? zORqW#1pB(wQA;&{YbWL^=lcbI6s2vm=vKoZi?ZFGQCIoFUk$E>o3Ce){0O(#8`gSd z-)Rhsiv@xU2^o!vPa%~q&B_d9mQG1$_V1jmqIGlkX~mLWPDQ=|jt&u{))v^kjKZl- zTUCTXXXw$7|Y~Eewu7MX4x`B1J6!UM}68P zHR6RB8#>L!CIthQdh+QCmk(};!;VcHqRe)FKxkA=zt*Q7?fU!t4XNv84$*&H!=2Rc z?+swmlQ3n01uRDPbE6n}OAP1yJ{(%!^?N@j;F;&6&nj41$0RMK4KDPAn|fHAT8v@4 zF?96Q(A3|<%%aI?b{WLnYHT5jSF|jBU0cLFYQ}wir6|XhTwC$-hFM6F{L?IhX=s*3 zUc;&`KHt9Xyx7(O$-aZ0Unl^tk0<_ZF|>{_D>`u~zTbf2M8j&L-izFt2M5FU+j|2w zMA<(#&0pDWzI|@E(73XZHjw8hd#JbkkagXZQcAs=&2QNAu_=k7@XcnPHBuq}BefRk zwP9+9sc@NG(GsAPDwUQ_nUZPYQ2HaY#|r`|SE{r{^1vsD^`Y{clkfL9zcf)kdQdaQ zBFF7#^qP(f&8SbQqn8!;{4cJvC{Fye)%n;YjAY2=%3D04yE_8Tp2fb5P_$h-r*nGx zGW@H=YQb$wLBz(AV<=CvxFXW_%!-Zg%aH~G-2FUO%8o3&l{Fee?YM}k-GiaWw+56F z8dR9Vc`7^Do|gH^mCwje@Lhn0-3_Uffok0A_R?-yoeV2BI)DMY7AQh&nJ4Z>#5ruq za7jUgf8@SiyRAi~TirH2!d7aCbZ}sBoPx8Pk zds1_g$x!B4sA#`$_O9dL0GzjEVmD?`s&?T+XdOb5@B3SRJd!@O>_=S_7P(mepx_A+ zm*S$C62M;XG@RQ=FP=--_3U*#zx~jI6G?+*``NA1|OM#declOvV_Z$trtL{E&!*wWU zAK;_*--~?Hm-_lGT}|a2+*l(2&dc#nYPTZT*dseiaxc>0qnb1xX?Tcsd-Rf{#4h^t zdO|7yulh56(i{0M;+obJ&fXjoq)V&ZL1yG>h6MF%eY{=mF9| zR|t>P3Xo)SA#7Qc6OaD^;2030t|ax4an`QE$``R*zq4GOUw&@A>_^&L!n6>g|4>msJPnJ&gWfJ;c}#g_Y0Sw%5?M2 zUN|zZY}!0H4W@jG@x9OJ&H=79wC|7-T}yMHw$CEK(>LBPM0uaxb4}(XQ6(y}@jrD9rOSj#uzwQfTR|&~zUdV9!_+W(}(Z`IZ2M^emJjqVE zs;8EEHblrfV_omG-0>>`g(_v<+67q#w!Bc5A=>^nRofj(XuCz?&eQ(lw7tg zU$m0+b{h<`%Ot?%Ek_^n$lOKB5B@Dp*(!M^OH)32^}~k7FHXQ+Q6o_tv0&qqMg?kB zs5UitUVP6~(t<(g51@Ghcw}IHVG@@y5!zJ^eAKo;>4 z87SPQtlr7`sE{b8fCY!Et5TP??Ka}l)_?xG1+HS`!ig8|ppRDB!%6mD^5E&AKi_0L z=T zUbe(R<=7?pBle@^Eho2uB06huxE-+5@h^xrjfMn+BzDC(gtdD!yK_|9?gEwFgjaR%rhb;|y z+14f?yBywGd>-s!5by#)c%+(42(H&jz((`2?3b<}+c|)Kw~|>Kp8zccb{2?s z*tGSxT$W?V6cf;8bI1Y^UrmivO7abm2oBSl!eAO%B3xWGFHViluc;0VBf$SXkKJ;d z>gU6I2h!Y90~QZN0Blg1P(&m0;C@rb_3@JX4d4}w9w;L!f+a=_&t-O5)0C5=9l$i|yU>n$fSI}r?Y zWho$)5B4L^lN%tIANF&3(7{=#ne{;7(PA&Q1q9w)itNHes z&p-x*uUR|L^$d`TR0q8q0OA}9_<{-T0crq%aB6z|5cC;9bh3d9rFUV|wVI6;&`%Eq zykObD((hg1a~n&`S3ZG3;^M$258ViB!m_?zCP?mxjL>?gRUJ4x8&i1F3w<8KkW@b4_4EsJIAD@(k6zcvuNR_g?uX3?8$}>4DATLheG> zhMp*-;xH)X>2=^ItwR=@8N0n0_f}Qq1`c!Cmy;5*67Tg0&Qi)=O*oTp1MVn5Yu=jw zg4WFqdg7z;^~3Q#tWrCIEf-MseCqr!0wPWVr~=9YHx*K71GpOk+;@g%>^`Z2##Jwp zC9-{o$#95Q({zC23;hNVoC=rfJw5U#^LP6K9NQ@h7*GsyP9XG3dJm@rD|QLUY;@4; zQntkUXeCDKI@3Y=OC;w(sRwsi3L>%*4Szlxc~zFAS_#+|Ykximq-($+fG9DChV&mh z3bJ~@^RL(C%1q5@q4-Qrx#k|ItruOP7fNY*ZY-?o(F~s+R{P zSJb{ge^td9v|IQ;{Ss-HWwCzySCe43I-=VSxIrB6;@&}Wi$TYE0v<0znlJEr!HiG+ zvPYj-E*YeCnKtYi-sn9!q5*j5u*)-*hAs`7W%#`Kmzb~?u&p&-VabvuCeOHzmrm0V2EM{Hmw=9hcr5hH2`{5uf3w) zkyE&OX^rAkH}dQwK?HKBRFww(A>ZC%F5JjQ@9Y=)KAM(TI}}&~ft6T7!u)X?3X1P< zv!F@`yiVs@Rj$Gj81KvQx}=(p;1P?}-yfhg5Jw6Z&_<2hGf_Zre!%~he)X}MzE9wm zd8;_sfL7F7IlS~Yp!i>PSeJs$E}@6WyI#y4UD4>+CGwr5)JVBLbiIMT><}{*6zMy9 za^H*j&&C;r=PEj(hz-P$B4qVBR-Dsst$(Dt!jSJx)y`khr)*D{ahPwKk1yfx{1!F> z-Eg(I_HfQ4k6zQGeuwT0z4!%u7yG;smid0*S!m1&7p1z1A)c^sE8-VnlmPp8Qtj<( z;&Mgff{I1z$q6QnCUxjdy$URi!(X{_u1MzG+kwKn6!L74Py1Qor7bb7Z`SE-yh7eU z`Ds>edz$=aErbU6q}s{$mD59IruYdo z%es%LDl~)gP2FS}b37iOy9iy#Z_T*$#i95?d2q{M@Y$EVvqv3!Fzz_$1R#WERe*Y2-8i0oL)6cT-;p?CMZ&h;bl*r zyh6R^8ily!98x09vU|0*iwe>`C#zi4K3r)gPmh9Ru9_#N1^r_aB|W$xb{ zWQ3HnCIb!z=^;-@bd~hYuGh>Tb$;HnPG>g`jD`9N#=7HLPj(beD4&Z#w?WH`tCW?d zM$bUP)ra~s0zP+ZLs(Ri84m!+M!(mX|MAJXSEk}nBZm@zZQW&EN@=3NF zi#@qaub&2%j4!4k_1s#S9_^TqCur&xUio>|e6&&;%W|vT zqW$VYBbDwe*UdH`Z1}|YzQiNCc}p^!h^4V0qAee+yx9@Z8@FJCX;OQW`` zjH)%rGJU*S#ta1@$++EoxhE@!@3Tw?b~AoNhQi!hdcM7D85HSMMo~3Mb8XX`X)+UU z^Om$&BMwG%5sIM3H4Df8vXNUMyVjl*wXSHvjYvnNWzb$P?6d2RWo;e*`+&=>?w*?Z zt_$z82 zzE4b0^yRjA#O4P}J6hPpVK{7(0nO-o(^W8EZu;^UT4^g#pzn1m{zC=z$;^Qnqm8G4 zHpIZY3a1s}x11ahy6vhtlxq|*Qc4NU!o+}L+|s{06p)@3hgp#V?n3|!-P%ox1kW3`+w!X_}kulJt}~SX8Zh9tHrzJSY8;^{jZipSJP2wnu};fGWUO%eYHMjzcdzi z6N2ui-A|X`O655%K4#*9&`{n#5O0RPX`y`GzhL2rsVRSxN3gDgRH$V!bN^gC%?(~5 zr)`OKjEJUFenbWFKlAVldnVIk; zVO(^||5L>OF6sYYa%_6sgC%|p<_^d2v_g>LJwTki=<#f>>v~yq+;wnFyBg0=2`Mh^ zcVgCOEbo!%AYC>MYRIQGg=DQMR6bPOA|4$PZ4(y-D4h@afkdFi7q-b1LkR8XEWkla z1MNlD^T42hwjr>Nl}*;ocfGZidiLQAR1gM=Jq-PSt^Lz7^4A~a0Z66sv_KJq{}_=} z%Vkp{&lG1RkEZAN7EYAeuwOV(A%rp&(#M&+V0fbP(L)m3W%3|#+8@rJX0CBe#~-|( zW!Nkdf6S_Wyj8zd)@-PAGHkClJV^m1Ji9S? z^QQ59z1ETLVoWMby{EM8Y>2mp#BLg%zyR6s%2B#aD|jnB?ff8U3nfZIZ(t6#;YlD3 zAuu#+tb0*`{%2mCBi+|%ZRh`BJt8SV+cBgBRk|u;#X{;BU1^+b|1Xn_t z_=;18Ct9TwZ0-LCN$~W3!~r(v_!)b6bya9YOns1NTt;Mgx0g$GRWPH<1k-LAOg)cy zKy8y-$2c#Y+gc}Zr~5d7G3y|Jpo+K{cKW{nelo#&9=VqSxZlKx|Q6$|B;FbdvA>|AI; zQQKz7s;bGhI3r|P=8XclWlsbLW#5NQjeJQ;T2pNagz2Vrr2yxNojg}nlZV#*O|PJ~ zbp{nCld%d(D9vfDZ~4E*@CG`zVHK&*2 z5|y4>qDBvmkvlcIzZ)zzc-o*Ru*V9RVydh0Ifl*2{4MZKwpqB+mjUBj&e;qNDuwoe}qho z6E*KN`86r5o$-kcOAn}7c97|sS~FzJRmCyzX$WQM~KHse_ zng$Y1iJyU zxFwYST&W}Tkdw1Y5+Z_5_6u2DdRgYg5GKyr!GW$SS)W{b`9E3Id!ULn@s3YQU-x4w zmfe8L`{~4FZkvidO)0LO67&BrCP%*hlxUmnK;`2$afX?oZ+nI&GiRU;beDx{BFyGf z0PI~R1GhUZq$_Aay~jF7&usVeO3uqWk4@A<_p#Emr{8Ma*S=SJ=gEuz1-Jh4O#lD@ literal 0 HcmV?d00001 diff --git a/images/arcmanov2.png b/images/arcmanov2.png new file mode 100644 index 0000000000000000000000000000000000000000..7f54c0d277aaa8a9e3e81c05947d13f536b18664 GIT binary patch literal 29362 zcmaI7byOTr&^9`_yE_3Af_rcaPOuaOap>gTCyB2|=RFi=TR0RX^|la*8j00amCz)>Q@f_EHc%EiG8yoI=don@*02E&<2?-TB2?=TyX9sgDTQdNVjm%C#Qc02{9DaQ48Vw{d4|~%e??O5Y zB)TM$OCf}3k#H8!)VG8XNJ+xuixVQ?E5LKSDInIu$5;969xo-PO{9rKkVLjv;aqc5 zUVZqYv390eW1#8Y4bfNwuq!K1Y>$l?0hCB9AqN?e-(+F$N z$%X;ESMJH0X#C4_O2{5ezmyb!1mQa*gF-yUdHr5BqidR30Cr4(oZ--+W7lCK1`~~l zc@%rV=EZxBr>y2Ab9eNEc|*3=FS!djp%i*H2F`aPAAguxn3h&NicX#TGqKPx*-fsz zdustTE-0AOD%a&9{F+)dTptw?XTeYJ-1PdVqlwh_=UMCP>G&92Por#n>dy_|H~4@6 z9AKE@e>*A&E0zG4XOW&cHGa1zQRQYShCQDHGZ2Wb0+a`R?Li#sLDz!u-h>qjU_*w@ z1c)Sq;d|iHsfmu^L-XK|5s<`qO5tMiNE!nh^3Zf)Aw7zYM5{1A-l)CJB0D>f1D=vGk-g3tPX zvQznCu+ac-XcQzdXFl>puxtmy(1Cdicgdrv=YlGrSVfKPvG-&DN?E2JnSJ;Ms&(Yn>@vtsefXw_yTx(5!I5s`GUWB zfB*bvb%lhE*gsHe7|o%SMyErR6 zW4U9EW8`n7k2o>|quWb-NV@E-IH_?J15KONo6?S4EoMRlH6gr%-kbdw%XgeUv|UL4 zSfVio?~w7-k@w&YJ~A88=YRh#-6HNG?;%5nyBqmVlFb;MBTh?7NU=jIi_VW`M)r{I zt12=>c4N3}*8C?g1uvO*Kaxkk{55nY^~7pT_sM?Bdior}TFCO38K1RLqExz6%~J9o z27>bE0xh)?PCvw4hVS)P|#SMimQX!HSmzQquElHwzqVA~iXhWiMp&^4CgF{)B>Tn@q5n+*8 zAw|)TA}l$B)SOhKF|E;oG4&DMg!CUp<2oZlBVR`JlYIwqjkFuKwKXdJ7K5^qvVQ2c z>2|Khuhw^b@5uG(@@aqKLI{GDd83;DF<;q8{PUL?Jr2z}jf+TJ{m7sARjw*lVpek6 zQv70`&CaQpmEY68H^|4zSB$A?)Rz}of42@kl;DXbm0?jjsa^4P(^^vz*6Eaf&Oxj+ zD#ZEeq-COoS7lX@Ub>veUoACTzc}{WdLd)p+`8Rr-^Sgld*<^z%HnbfR~}RYqQa}+ z{Mr3uT=9NEY7OM)E&bQ*Ni#?-)!~n8iYxbj?nm7`zu6<$`Has;;_d(0xBRJ{K95SM z!PbkB3m*Sjal!B4MR*7HPSCvs|Ub#?dQl+fZT{qLTwd}gY zyqvLO-*D0-aZG;l{Uq$z>c;P8;bxlRFGT^l4h11O21T>LjX-|3ZuSj-jNm6hm+Z`J zBHd`cU#(Z|iMqbJ6?z6OW$hQMUaLm#M7PhU$`Gb1W!;Tk7@;iT8*d?~`h~(p$G^~v z%d7GWuY;=n>?7e#iQDDlxov1KWT<(mC%D% zmgfDtQbA&}W2#mBj+S1({y>bmH(_r=-7dnz8!WXSCQkc%c2ju`SQ zq&mbigeQc>-f3O6-7L^%TzJHn!k$1MM})|T@-rm|o)x7wB|n)NWs|5i>k4NZre8+hH1mVKd;I>x%j{)q~zUuGTarJap9MF+We#l~Q# zlOai{MfM6ON+!sZL%psGS5&j~T+E!-oFe8A7CGa%Q}Pp}6JMrT$8tuUM(#bCteMh^Q zBNwA!Jy8R%)7g^9xG()!=(x*R$xZv&S*4){zGjDL?S0RQeDGLXa7ge0(!GYg##%T& z)`@VVDute~j)~fwK9j~*^>$ajr}Gh<3R*k2S+}=txp70}zUpUM8_RjCDxT;r%sWft z(yv2-x%{HMymH*h+`jxd9$EfYuckRx{QG+&TaLeOCKc()Oou{tzO%zwL+e)Qe?JoIDZBZK(^TIy8yB`hWw3UaY zPNIg{M)k^DnVnmYQMcrKlNT!ZHHo}`c<17`NS><&SK7Z)8lb`vKEj znU;23}ole!guPwQU)sWz#Q5OhCdhyg{IWrsf6%TtF4D`Dyv^(E0iBFclGq zM*+O%UJ-t4W~5VMXv%`XjM_w2RS^KZ-T^>B2mm~Qw*vkFfEzmi92x?EKpFrLIcBwe z76KojILd0f003qGe;=5R)YKKQJs>A3uI`a_yrx}W*p>*0S$aQia6XF@q#Db^zCI3jIEIzn{IZSE z&+d_!kHb6n{w$=%+ArtJE3)#@e{^pkw76 z_dZB4n?BP%JwfP7nX5szxF+Tp@}zQNQ?dAp5C9=FS{abmt2 zDwwI`?|$~xJ9Q(}YRO$`F4Wl6>BgOkaeo@>X5F%!ooMRRAs9(yfzZ@j7n~tFh?hF> z>_vMK9u(z>*Q@EbKH|R8tJ(*-8gqp959fBj{#0&`EBWQ$IwR!vW_NA3#(0W%p^+_# zX{?PsGxlqHP|!+J`?Kx#X4g&vk&CwkJAF!xq?z8K z>6g>}eWqAuw-t$Dx(igFf)5cj^+(tq%bISoUO%Cdn+Z(O{`4X~mXAcFlY%fy+5aB) zl;KKzL}?j{418-oY8RS2#2)(O`)ThdPGcg=^Wz^#Y;5@GuP{&Rzxt2WF0tPCY5iH< z-2Un8mM;NLBH~f z=IYmW49Z?`Nw%N=Tpmxa@9pd=W<3hc7`!+wr0UeM37({CtGzf*9hkdD#9rO);d#F4 z#_qa#oU&(}p%5LBiP?LW{;T7fy;~BP8#E>Rpe?Ldt=;8Zg7%P~-I;Zr z)_h7N@5aU1;>nHM?a>kJ?LVPki#90Rh&wvn~YI6{~e~6KXnc(mmtR(;`CcF&0h)n32i$iq(mAhoAkKD7k(YZIJ3`g|ew8ZXw z=h5tJ;Ptrd`&OT{&(m+)V>kAGS*?#UCJo_hQh986{}v>ulIVknKkZqBXS%OFANgN{ z-}e-24*Bb=yW5y->Z93ru8((~kB^qF&y9B{Iym78A4mI==+yN#^hgzdIQ5O}%a|xH z1;xxx8=h0h;n)eNO6TTgXEya_I`4MD6h!LY`n@&Z+g1PRc}RaL4e1dyH5#kcridf3 z(Z(V`n*MkoK@1M379lg~a%m#)E|mo6I6JuEuzk#Y+lIJKhX&0VjE-bxlI*hm+`3V= z%u7VW-`-n$B0mw7t=K!T8U?3qUl*jVMkZw8psLza02}|QCz7#*bkgCc%_f4l=Kk}+ z0l}uk1v@{{jb#YS(#ZQX0r>8TzKmHH10whUnZCP{Kf~KA8>(X{cBmPlo~|V486BSQ z2`hf*vh<)uHbj$6dn4$xYE>I&$@e}9wH~S6uVQ$2h~o_DiZ}HJ9z$jyXgselNOxX; zJB!<_zPPZ#e=ea3oYwpuQqQ~Fw=P-WP@!e`!hQdy^Vt!FBW7Hqnd5JwnvOD-I%%Y za1-$?l&{wp{JiG_2`|L=P+JeuV;w3Uz#j^wn*2Hys%#x_Md9B_?2THEG4=9K0x~P; z^!F!kpJFeY)qB0ae_nl;`|qw|d~&|E?i_2l(zl7UmjzS>e~WQO$WGw)GF~pn_4Oma zp)8pC7C@Eyp);YSHPqFcZ1w5U1wl|FT9-Ael77t7DAP?7GxyA+dSz1)TCnKr32ZzR zc49JV1}A%{=h{=%*u8%I>G=T|$JKj%a?FIV3+<0HD~vr0<%+i6rlMs&|8Ck33=UuK z^b&$N@tqD&B}cTrs{8wJvkES(2y51MxGc~5yF%TSDFPpNI~pc8n|gT`<`I!tb-kVN z^*lDkj5ocuz8+99jfb-PIS7irIzJ6Fv&=7_7}#@9|Dq+C^KVrmaXjA9Of@5?q%vK{ zIOlH?=iVV1br$$8(kewn`H!+CN(0TCoFvbfYFc(ozplJ2-)Z|-!L#q5esFH}S*IDm zw`4C=#tV93E#;HkjO@co3cdzpkawTHxx@Q36N;p28YvW71K2&hP{{S$N}WDtj@47- z@fk1IDk;j4%KMGVf-m8kq3f1ln6Huo7oo&xTpvVv4Qy&K;NCHY@YS;&YhI1Aw4L|- zGv5@z?Skb#+F9`2uWpiFVb9~*`6-YY$6J)OT!4l}*M=8_YT#c#My66HrcDg)u>s#h z(-PppFDk1z<_@a@&6xc0e~JdQ=|Xxu2ZAIftPc{|KFx} z8xVw%ytBueC{U~saac`Me6B1aK+1t0T=&1iOO2Q(O2?)+07LcX0GtGLVBQYh+}O;B z5GKVD9(W%CzRgEyDLva3SMi|F2IP^_kN}${?@f$`*v;0>xL=CSyq{$OF-Pz=wNw)G z#0oolrm$>VG7xYl$qDr6L@-aZfzhh`Ox6GoIFDood-xrMt)15Di8Nz=Tn}lW`wat# zxbr8%06N(3(HdHd*Yxk4pQM0*)JRhxAQyyV0yZl!z+1dg02qetEc3$oiL4C-1V#Mc zr~es)@!kI#BTVb44>*=H=yc9NsJ%U1>R`n`{(4ZMZYmUTmK7y=M2lVS;h=MTr#YOz zyo06AY`dEHUfpQXuPes>rEaHF%YQHb_v!H)3_sL$^^JUF4nlLGyr}WcE)#I+)fhX$ zdj!cMZt}eSu=_S8eP5CUjSF33+-8`OByz$>;r;>6Zk!z=nQ|wW|rFSPK zcRLk&fNQ?`(p3EBg<-QE+r&uvh!?Hp5(rQQRnUshwqP04N=qi<$R_nPy%y7Z&_{`RyMi49VC)WB5~*WJEeHpxOxHT0M|=4F!^el$;M z#`bql+17r`(8fJhPVw*$?Ixb`FpFq+JLF$wHiqme{xNkLilr0obg_x^ppxfK_)fsN zPM(K#xBY2f_GPYcA`)y&D2&I*o-xL7F8|%m+=ox?IR~}KlBcPnZ-;dJ_g3x>1&f5Q z*KBa6LpS8_cgj9eCOz$&H|S!Otn$ATK!G95iTz4xN83ku%+xw0Wgd5RsjsQlk)fZ1 z)_u3in#JgE9N|OFAay4^`;|*b$H|CjEw-&3)fZW?zd~q`+^9c zP|UYIjLp)=6RfL`tSK;Z9@YdOlg)@g1%z=04VlqZx`WjR*N5AFV10jPp zV`IhTj>~9P9>!Tx&P0hysuFfD#^p^^4NxnKSM$DVh*vhiiuBHSk%=S@Xm#IC8Lv9; zby#E5iZ^7g6tp;*QFG*?Pi`@oKIz*%ogJT$h5gs|=R~Fydu!$;-RMSO-Nsw^NpIa& ze@X?Zmb0}w+|R%9gkCQQU}@LASa%4;$_N%j@SbX-KUYbRD_ydZ&6$UC={}iubEdp< zY7+;L!lQd(qF3-D9AnVzW*kE#`|u9Th}2YBG1b zr3n%9j=Jo7>{pJ&BC;};Gk-F5w6i^4pF<1z%mB7_?F)UIP_-3HmOhv50VRgz?^l6z zXJ?x0mj+W^VXG$&2s}&o&E*ULqN!AYIA02F0r)C9x>LFzYZ2L}%*? zeg1S4^1E#ULl@llg~7a7WIkillA)zl_hqr$nT%`fFRu)IPO5Kbbs0l;zlZ z%xh+{G2Y2{Q#z0ODFA%mN={nlP(}4>$2Y3he_Gge^$v)>JUwb^?(C+SUfjyCb&xJ< zFG%3Wg;Qb!vos$QJfdu(mo-U`D=7B_{?7ATdBtu z`$Eo?zuDPdu0;0V)e?mn*wEKjqr&^hOpE{#!va(uCt*13F9r^DWU$OTU$U?TeGR*w5B!x4 zbg_IDZ+T_Y-GC4YKlU4ql*T;vGu#v>Gksyeun*b zW0FJvKN`F=lxip|B6YrcwBEavjKyocbo}X1J5Gw(5M>taPIF$x^mBq99VOD|#6HU0 zU7oJ4?oH+bdAdjDG@X6j%SxTUS9N83vsfgCPK#tDn$$LYy>@aXNwxhN!+1%vyI@5mvQ=H*x8Qjf*vRxGFNlspXIxI)X#D}A5HZokS->_ zSBvr6cAum8D_%`-tP_wl{tQQh26sN(RPx-By_8AvV<2XK>`T^KpMK^8v|SV}tLyAQ zRX!8E$@c6D=T*z#*=hN3H#nC}vp56r8`B!!udH5=WoIJ-m3hi{PnB3F0 z7j4wp`T@g)Ms21+_wfr%od8=8NtOz=eho14EQrm1q8=$Jp-q5EvN!D{ZECn8j|~JRL-`b6EUY zGY6HHv&Rv-T=G=q+i%myMZL;}j)Ko6J#V`FUIixFDS#j{vdYfA8aW z%0^hiFe~pvDwc^c>Yh3GrFqW>JE8F#GjM{T(9chku)6Pu#&<=DSr&KV_gAO?|5*b{Q?9<1p*mU|%v{p_wxy zb1;8gwiGcw?fsL3TPG^tL(xdXA0)U#!R*r}psE91E02$Dezg4A{MuLa|3%v%l;=~# zLIlKo(L=TnzU(9Ly&3s_3jn59AEW7`P3h3!rsWMLp7!bAPDwb?fN4_bfA?*(z@Uf1 z@pUq-NI*q5Fn5z05F=lpGdOi!jkOuxW1Zpcc3#XZs%J%)Z07$q6Ce$yPE*};*o;L6 zJ9HUMjHnyyq|+d1bFZ=KQXsQ1RPNPIOVj2av6*P7toe3N6fg@y?L2V=3nRj4v^?W} z_K9G(Q~|MrLPNk~(kb_JujR25tMb+DCEbna@@h1bpS$>5BCB67c; zs){?OCn|4;1G<`srgf3M$Qqj?CMH$B5F(N~zj<~3~=J&=axr6MgT0 zWafHWo^{ZoSbsamJZ3)frXqV0c+na=YJ&j+&~6BM?+*+`rZgGSf@xSp;$KGLj!cqw z0|&K;OMhnu8>kjyZH+hmUC>k=y0h)NIPJT^>dL&V^Zb*}X~b7Vwl$ACALSyPYyl-5 zX_8_1D)`>i*a-7x_n<$=Ib^OmMSo8?)gK>daezvRu62JN>!a~o`t#Af=DrILT0(SS zyR}~=aA)6)w^GDKh**AkvYCqCKhZlZPfqk>b~^rL*k)s`9l*5+SA0G55=^}Xi2!Ce zv$yNh;8o(jKot?@W2l4^^I4$pKN*++!NoVXN%{w~51Jod{qE~%#-Xxa3AnEl>l4>#7%Xnm_ZYrMwDm_t@htcMPTkWGOp!6GL-(Cd`lAgu=G>sj&QZ@7jkYu>3Gv zv~GWZG-)a}i_vsQSm5g=TDP~)jT&mY>fIF$1=T28s^zwb`Ta~%)k5L69vtw_cLS>* zBo9eduQt!h3=&G)OH3!d|8gw%LPhzGkMa|DUS*djJI|(?6L3-OZaI(iMCGiy`r$fvSVKKCYActb0HD z3ft#ma{cNThmHJNShH@?AaW@ba#z18DX50f{`haziaM#~uB1kH`|aPs0ZCJb9p{x- zGEr(+Z$!PzKQ*U`&Mt5ozd8$8E$O^r=D3E2)La_#`rob?MKkzWJ69F&U;&Qcrj=yT zK!5)uQOHiKYC~U!buXjy=gdjX_UG^p?<5amlk&~KguYKO!0C_Bz{JKM{^^xSgRXD) zjqOuB9hPIA@@wqVJd(cO`G`rP+5~GQ$uQ$)gkqS zok&P(X4VSyijp8{y!7nvHz^8#AV9Zr3eC)tAAxIGx2d@@yt#AsWi>(2`(fZsTL*32 zi`N*D1EB--h)O@i-S6Sqm*F%F0T@vtL`0;3 zOH4O@CiS;`tKEsh5~7FX-+eQv{;w|qFvl_@*IP-@rZusLW?byw_Oy0z_<2{)jvb1xpns$rEjhsuHLAP3|EZ!+BG>4TC7Y9jTs| zQ9&{}KuxVpj4Y12_a;g=x@2IxBR*&=pdSH@hji!`8wkuL$1)>c3FLu zXo|-Y5xvdxTEotWpFMkLX)c!0C;m85_rD%&H<*yCxg(R7yOLlkmfb}-U}Q>*7`c~g zRPZ|)PD6A2r2Z(8zvxo$l`NyFhL)j@1kV1k6x6#7R9l$J4j?<7Xq?77%5z&5Iruow85WxGDTt7qSYF7O%dKz_33R6qmtB2rO)_ZhULKFyS$#dQseb-vQxPyN zGe?XJ2e5O}qO6;sBrZ1XLfH8f}`T z=YD(Wf*QI^`=qC(O@FsD7x2h9LcJT(y06UCd4usAU7Ofrdb@6aVD9B9D#}6emF`F+0CSmK~%UTyDTEAvDmzR&=CFjqw0W3aUisG$F@< z$H6hYr_#n-Q*qEtc=id_I$5WmJlbcb6Mp``h?6LkJrNaQO8j!~aQ(yog;pHr=sNah z?AObqK6lp3==j@o2gkGh)1}ouFmJK1PsHfGwGZi8PX^{q2%Vq0md&&R!a!q@`Ati$ zd`czX*B)Usr!oH=y`#z^ePw$8x46@t>x^a43*SjGCDlv;RIx|#QtS=Y zjB4BDzMsTcK+kEmD={>G&D!Nn$<%xL!ax0;4ULD#ig7|9s`-9G$e(KRDb8KQGv6ah z(az7-f#`;rlR5IgMd5nu|KMU#Q*k=GrBhP$VgD^bYA0;U%~1?f^UbuCJoL5d5EwIM zMi)9At|;5@m0b5QUTfffPFXMX&6ggiWJ}w*r-+)SlI%Ii71HRK6kWfs>cpadxoVo$ zUHYi)8FF=WxZ-!R`(@%2EKq7xCM0Yn`OT%Jm$n|Ov+VN#UNc)Ksk)JB_4k_f@9r@V zRno<)(|=!Hs{H+N^6+=~#6hPMejT8pxo|3N)}Ob0}~AI68wb+_CRs8t0zkM#AZfkh~jDa6b5Kwn%A;Cy&A zVCzI$jU6tjM=)^{qrdl46y|X!!sh%prCqN@2bD{uB11MM*v+>GXWPq9e_dj(^vW;BOE;YazkY z&DNN9q~^2?csktC^ITNYqpMME|H-ctnmlYnmH*FKfXs0t?0%qEYNGE6U03ph(;5k?X*VQ!>clN4)_k~398K+BVwEOGu0&QOR#;M+CB@z(<1+#!eSunuNjPTN z-t(5*jb^^wTA%GtMyiyA9dI;}$)O}Wi8dPF^1f3pjPJ%!CPcnU|J$CG&E6D_T zo?eAe`W8n@(2M7KUBg$D^}aQ^)Z^ zJgZ91o&<^7kAO(ou$5>ocSf6)V!`<;R3CatLg~$qiZ88+Q5g-sQf9r&o{X${c%v%K z-gdq%j|C4)=*VoV(S_FreG6-~&Q90^(>8*S#Cy7XCSIAj+QPXRSI>?%n}Rgy)@!yk zpE`gXPYkuyFZK6^h#qQr%;@Tyxl*1#dG@_Iwws(YbUSJ7BT&R}E`LE_B(;CKPiuXM zB;I5J&mnlTp`Fu7;Cw_PjB^fuF)Pxy%#sZ&7vjR+JNp&Plc1jk-@!W2#;%fV{)2P! z)C1!7yrSnkZwkW7;K5AXm|5p#(mYKe~hMua~q z|6rO25P@&~Wgld+D|xeLR1}R4NRe*P%&WgED9c9ZBIN!r^H8%`^N zJ;Zed(a>8O~%5ewTw{~KtbJN$|7Wm9#buIT*VBg76)lssGBKS$f7M>GA3-^yZ4ek7ougDL;b z?=eK`sl=)sWurSLMf$*!n%2117puS?aC>KZ2Ft2&v!-`6IB;SxjVhUh6>I$RVwm;p zj-L{aCORkHWF$nupQDh(14M<@~^DnY4zSC*3Jt+D@;_5H%$L{A};k22*8EeSQ7 zHgU^oZpl$t1NxJ?em4~$R6Jr{H^2$^1`N6UgRUzfJYC2Qo&A=9#K86g>!6Bbrk4F^ zpz>%R&>;!x*|JWt@b#}_8vhBOV=^ zefg=;1J%pE7ceVVT{~;pssYh`Apt-^mcj4FJDeytXUKC?I)o zkD!cjDy$(LZ6HrY?SWQ-vhWxa<*XVJ7fk5CX~J(O#)Rv!Fh-ZGvf!J@`arP+j-l+s zHeAe+#EQ4OTlNe+kfr%8e4s#_&C4Lb0&H~A_&iov|5F*T%*r9yjAuqCzYuy#zrp3$ zdkgcszI>(E1ML)W8tD_m#gi7Sn8>u0+lv7~M~|;BQQ{>; zzhS|Iy%>lOJ-&{t%Gkvi7TOY>vuG19X?%qmz*xWoEF_L1+Lc~6z zej^~Sq6w$cxg>xKRIqu?o+S8_0hcHO2DwWlF;F^8lQ>^5@yFCC;G;UfBnfb*nf)WN ze3fAn7Zhv`e3WF7Vh`&F#57n8P2ju(h5zT+lm&N6M0&w=*V)lA8T=+wU}mx>npSep zFu6xe9s7B(i~$f(GoKy{LkbddUi4rQ3`#S7T)=J?194uS5b^!oT@Yg=3Q(%I`;6^& zMo|kIl8n*e_~^F+QTi4;phEvF_nS_en9rEo7KRkr0MvMd?WuG&tgp7m0_HzLP>o8} zKJS1!m4yJ5>O5y-FPu@IYc6Z2fdi^Z2!u1CDGu!7>^I62c{Tvg7i^dN@F}c$BOSlz zLu-i8f#>+^e5y0+<5!)oMvVi<<3!Z2?Y7PXWiKA1fd7Iz)yJ%!!?2IW$_P*vLHhmG*H#lGE`^lbu1bAi{B}#eaGF z1&&9zR*!tE3$@T8*jm7fX~RYkzWC_sn1p(`8InY$u6tR`VC z&A{i4gJSr%h-OxKhO$NUiTymenl7n9*%5iyDf&hL~V_it8fh<%j67L(4(O zZ>r!O;3h{5DCVA5UYyGL2qq!|ydn*OJk{HFs)uxDsWDr06L>Ba3T@&$T1&VIMU!LI zvr!(f#aDS7K~BtYl9$82U-_!+Dzw_fTJm;lWF8cpfJzCYaavv(UVsF+2v4l1Q%S(V zsMXhC;(76*|0fGPqbJ*{NmiMf&C;$!O4ojSLPIM9moB z_^jWc!D)^{0KkXNWLkT5Vmt^mAc&nToLd;Y@qwq)-1`F8S1eX3l7v^0|$ z{WtMvG(bUK|1ayfW16S&?q>4B#=gNGXaKNq)Y6-YfU0q(%>`Z$wQM5VEP1j z38g!weRUU}J|Fx{-D0oZln!`}-Vw9uU|2m6)1n{hQ*cYSrzv;6=%YMCG|0M~fgA8kt@p*T%I(x4-;7WPDsDoa&x`(wA7_Bem&gH|??;gIt&^NAg zq3sHA5K5n6|1S8huWQGL-g0a-s`#17g0O&-FsztoMfhU@gO5`-AojUXH?USIIy9Y` zQg^{gy~k`K;X5l{KD-O4n;3QpO?f9~9U>%{ zodE?P1tYT0q!w|Do$Y9&gSyfC! z4WiwM#jk6IlXIJ_5;yO}=p=f>@=u?lJX>B>7?qwG5qwjV4HFiP4J)GeIuJ?)hF@H0UqzL!0>#X7Qda5e3mvm{Q{_QeqFpF;57|Tb^vem#%x^gkL#)`9ND{?mHHG7+7Yo1l{ zpc5B+-XsoQ#+m21L?`Ydr5xztfyTn56A`9)$QeSLO#4H96Z@*C&u4=0tYUIu;)fXD zp_c5liTQ{N*BleAyssvF=stfzcW|*ixw0O5tQ&9iO{VS1oNsH~nK6{6gEOnsI`f~` zIWKsB)R!ir_|`gCMe#bM&@l#_)mASu?i|+=%5}F3#RX4^7>0iv(Ju(x`29OJnM2U( z;-FbmbaP2g73`YImvB^6eIhe($6H2?+ZTEZ`GnrX@oidwF+`CIy;EW zMCT4|(v2{TLbZqx5A$cdH1pCJl;rIHTl4V|fil@sedO@}0x83R7MWpcP+U)S3L($%cr2HZ>@}58d=282!1g0mESIdOtxUi+ang z@zQ8b5e%F#b`!uMgHXnqte?ncd_O)BZRtrNVqSp&62RjHQ~IalOuPcAV{;(7uugtf z9JOOd$qwAFQj-^mdr)?$DSOd@^D3qa=;Ob2hF5QSpX@pdIJ3PZ{2KnBm0@XZq{Er$?thdaxlaM8@+nwluV80 zWqKqot552ARrkj8IS^(!WRvHCo_AI2vp%No1fSN#$%p!YTu(2fwMhhO{mDTEQ9Y02 z0u$_jZ%0yisoL9KqOrfv>)~4?|!=# ziFk`zf7a}X_ve3U9X_ho=_5)ON(N&emfZ%Xb#Rf+Hc7-LxZoW)5Ufm=HoLU6Z$MsSbn?YJ* z)zLS_T5#~%$a_mP99M=T3rpkgsHuC=d!$T=Y10D#fW4+0r)q9MQIm=z@F$8bYh^U(Wwd&Ex z-j9Qi*|GnX)kh``mYj(dfr~WENdL0yESJ%9Bb5Nc8jG%;m*j{U5h}dTqIsXzsH-kD{(GPOzw{p_+43J#)OoWj$&*izc#)%kRS?suCli3iY)Vv+CiWICl^UpJMQJYPm_-+3OQhmXp0+nv?xseFQoU^gqR4ivzt2$Gi;d#1xVr zJyLeWTEY0j@Fg7PQs6e0jsRQtvzvm!MnQ*Y8A`rx5J^jQ%$JumRMX6USPKssTt_`Y zhBTRnUQ{9D2e>kcwQGXE`zc0q>?7ShU=25v$g0%(4IT_|65sZ$XbC?G4mxUu1er_c z*knplAw4ik9Us7$eR}>Gh>TqijivJ#{13hpR)IFhIW2IjV zP4A{Lg~A=a%VSu_gC7JJHp@=T+>;#*a2GBCLl1iH>Z}gFF-SdO5C9CTt|KDkc#YeLz6Ll$fA*_>kxExH+bG&&M z>iT*g2ZsxXo;Pj4W+9b!nx~(p13i1I#+!+s;Z}mV^X%ybAz3@(ebVZ$i6I5lK_t|f z{(rj6fOAOOL0YZT!uN0ufRr;=f}#*1OTJ#8uQ)^uHp2~s38x#) z=;W_BhThMqwR@esqX-|!5x2pe&-1B$L_G;Gtl55Gd{9m3V*~N@F~a-E8bA0KH65zK zCq!jm`Oz!J0?johi@hy&sTp_`)LwK`wcBUTFl(zv;)=sGqIc@O0o`B{y*g^qXOV8I zbWL)9UqN`*Q-4t$E{~ze?Ok@tFL!a7>66YyQe+7b;3C+GzzMn!9P)MarTU&y2NW+r zB>lRCbO~`eu>Tx@2YtUsY8(KqdL6?@9wJn?&zcG!-qOy8tHAf7ud2%6b^oeH8e5yA zep}p(^mm#94y>fpuN)6AyJd>HR#z$1k{0mx0U z%~DlUWUaT|X_WmnFdS~JWAA6TWwj)#7z$LKcimB6{jF3|w(T2|0)}@xsQYJ4+Ps8z*j05IA}2N`hc^ zeH-hdP(csW&L-l!6s#>3LEw7$Yo_c*r3x_Sm1QtcQ*iB0iQMbu9SLX?Q_t9UU)HlIht|LZjSUHbqn+<$4TGj%9?P2@@2X+H1UWb7S65?|Qg zEoF66q|wRqZ$7Jh&SwQvR82jKo1f`$T9(+H;K9VxYyg z_m^j6e^@JRr~=n8*xG1o<3e0HrW#eq7BvW5=Vwf|d=oLfQzb~)3R^ZgcxdUbL5}eo_|Kr{8VoZbFuKW{85KG9<8r&^jcy1F;Ci zcPt5KdCkUE+j!V0;r|SG*Q`!w!e-;p|GL=!pX%N!s_HNL7yZ&LDcubco07&&cZhU% zN=Y|J2!b?{A_$TKQi^ngu;~(%5&;S6&a?RaAI^C>_u)R=F)k16G4|@Y)|zX^XU@$P z{}QGH;!WqqU*3^J(eeZJS=65K{ESg&?Zlv~V`^U?uV}+eHLqbU7f<2Riqg5O?c=klxwkz1cA27vHl7i0tfj)9TBhH+z8>ian6uu2CM zNhEN0_|({NjIz~VZxJm8yy|sxwq8ksQQ8jHkKYKlXjD4qwHB!~S_$FxfW!)mWOdP- zcSd)UsuAX>S@m0&O|UreNlz}De4+9V@$`c9fAkogxDqn3NpI6-&b70_W?s1>9wJH~ zSA`S{wCvT#?M{E2&{yMLqFYRIfr_?40rskqj(cU*@A9O+JUR1kD;qt-dF>g%g<$k{ zBH;Za_}KdW;bGSs!HErGr?GF4Qx_UqTR9Nu96swvnXmE{U0(C5R(m8 zYi-~2n$29d=DMscm@6J$x|2(@;)w$tBY1MKNJfLtwR9#TgR!SkCtnU&4dla&uhzEi z7sp{H@a-d%PgHP;}wY>L*B%r#&^!qB`?0}ND4icB(o1WpWno-=#HUwKz#GI z%-1v9k@2yHAMEI^xsX5vi-jI`TRDkU{(?^!FJvl!0bx$nGr$M!9I@fKSHS*~V%eC& zIY%Nj@iZ47rc;#k7eo%UHsAPIhPC%K?-c$V9L`OSz~Z}K#(~+>+~AMe2=i@-i??zno|=Rq@SszSn&L2>CWt; z#-KaQdO;l`=!oj`a9hjRy%46e)i()W5nW71t=W{m)ZedZ%vq!|U2{STBf8EdpL`@st6kJv=-#x$PBZ0;Zh&l!4zD5 zcdsL>^4tb8&eD=y_7yM8J@|oq#wP6p2JpI@5xQ(e(!eBb%S@8bk>lU5DLq>#!)ADK zyH?Wcp6S@H8w(SC7v*nN``gZb{J`(ApG?n2->l;F2p#>g)x*Avmx6N<6v?=e&z|Ry z^Pzciuu?m>*RbaH|7aE`ckf@J*rXqd;DUVVw*@>{OxuDQzY7Q4^YsK#1LQ# zbbMN5@p7z>lW$_$>J#I5P>A$v8;7Bs$%+U(MyWvGP&4tH?&iJ~iW(r=iXcl0hY)J^ z{O}ws?xqXFM~zt$4DATNxBU~SrXhxme!OgDTx&s<@a=we!F50fNvq*`p}0&ovYuV; zMJe!ywwQcwyE~%O*#ZjLK`r01B=6%8a7;eLQf2RG8Oy-Ik!HZS$VgjrjEnOO8@e~p zku{f!hbca65;^T}-Rh|*ajjR(hCUgEBOU>xj;8FT#9^YqZI@i4blk4ZEU$Z;STf5zPDv)Q!+B?T zdAOW`Qb09`uL;}}I{#Ei?Vdif`9ANek&=eLPF#>9t90xxOrORRIKYq+{#r6Qbng0L zdfyXjJWg+L_Y9G2sog6U!_Ov3MQ+fRRPC=@dTOet{cO1F6nJI}q z;#lo5UIFPb+!wuFjViajO+5so{48{nZeg4CuusTu6gCi_#yIGY{Y=!EH3kktS=zoj zn=d|kB3M=5dZF13Q%Z);fZYxrk)1HmOW`Zm_`)hz8tD;4MeD-ODYs zRVB56t=sxgK1&U_p3(;SUlRq=P@v`P$eq5N)I{x`?6hNle0KD*2^AV36BO$wX(QwYx@Tebv`W8kx5fi@;Ak(I0;y{);@OTMPUd9?k+U<;vz z642dWte?x7TFJmhIE6EYU9F7F$&wjPu=qysgSr$pr_RaWcZpMffe(B5mwRL+p9Gs# z(d{}7nfuu)RkVTPuySO^^l%IW{ZC#1icZ|1y_28ZV0g(RtH_uq*3_h>qrhq^y~WXC z@KC$AQOBxMMCqr=>-|=J*Y1cvLpbeb%-x#Hmf&E!jQ?D(@!KQORz*mT@3BI_qCPox z*KjiQ6uteEJcmK)^v!A-D!-^CX`h!$%z(#8|{c9OL>9)HdpAkGa zk(v3J5Id`bteUcnr-Br>o| zpp)vB2mzNnBjvr~JOB8HR0|VZ7ud0s<)NDRnkcFxPjP|KCEkHAA}rw$@*XVuJGd_pzG3_K{;n;#)!=I<*U`G`3a>uk`8zT%NvIeM4> zj$}`0bXNi)JpL;U#@YbX(alrUQt{@tEXE`eX?f~5FlN;_r6xof2+A{WT|K#z?nV_L z7?Ow=ZQ(mZZ}+NMk_euw2o!ws*)s8uj?zPswdPu}J!!nm0=W($LyU6V?lInM8=&_c z0>i(^?u`BHUv8E%%V0WAD^qf%539%Rf8u0o_V=LF#tfBs2cg{qq-A39Zp+K|FKr%f z`>G7w;~0*`|x#9ChI>I zF@Oo&S`Ax_{^L6-gU@2T``w{e7+YeNs-9faQcf*4mmk~LiVNW}MRd7I;>x}=AF{q_b$04@ZD~h-`a7(Phsq(^ z0=J-LrOPDhx=BZ@|JIp0M7;2PIKzsPm3`j8qX3ZES zu;NHRR{LTDZVV(F<9B4|i$lqBSmCO6jy0+jH(S49P**@fGN-5OP^Hbm?`E`Jsi_q! zQiTjcb&l17fMWxtn&9>ficS+7b*s`i3v>r1TMlx8S@x_NSVuB3MQKxxEJtRL#4+;c zf89|VeK1xhob;9S=_7070GPES6}B(qy;wMP*6NpC+_)TCdUd{|$UPbLbYh^)S5r?} zF**R)GU$*utad41!Su(d^*`g(|GPBEl$YRcDRx~mUBO2sDAVN7?c?#5<4(;!MRNs) zFA-_iZokfWZ0z!b_Gj-&Kfx*b#tQb9LV{0%!W_^}x*NqG?hF(go#BKKBH&){f2AM6 z>_|j1H?O41cFV<6DA1(TKleAm+>MY>X6wEvu+cs(cSO@nP~dqXMg+mgLVi_DU)~a5 zsdd&&aUs^zHj(q>8QwF}efC@`YTvuWE~XwmtNBZVbjy;76q1vD`-88;bwM}y7Hu#i zElLzqi^~B+I#`I>uJV@yy-PJt70QsLkb9m5x)R7op(Dp#6GG#?W1@`?{KG<*@1jJ> z9iQCD?;!sZ@ompI=j^QL_J2*yJ^j$zX_>PVH;q&2RcHP0#n~v)gTXDC?i4TXKbfUd zz;iI#`=E!XzAPR`N@i7%S2pa2ZMQ*r1=^!kl^;hJYe?0APGwf7Yd!Jy3Q=k$FZ)=A zmBe76Hf`7p3>aI^CnCy8zK@yx4oriOskKeqQ$=>{+Mag{4s`$j* zbx*&u_3z%>mSq&Qj;jQW;1)Mp0WPC@-^;hf7wMz7YL35;x1=e7r@myWf8fmAFnDLU zTLevka#KCZYyyP+uHNRvGbqk07$rWp;-N)-wcA#9xzUbH7k4zcHT~s_Vj)Qi-_sLm zL*QdAasWcB4>g)$5Lx+t`)Vj{Gy@-xe`r`(7}m(oPcbWYeAFjm%_268CxvLJwkgKcq4WXY@42A~XV{!orO+d0#>&pr9 z2z>RLHoGeLT!~NLbmo-f#((_~-)3?OlY z`8@AE0Z|-p1@b~!zp+m_zTK=-&1MI>o#1O@2F08GD2hyB4M0BDp1unEA&&w z(VR-(TZOfx#+7jfqIKkiv`u|ec6p!~{Q+PPOu2>qh96&RV?j^{0MNKUof}=@wpFYB z%+3o1jk;%YGp0VN)o6XS0lKq(@pY0L_f?s9I}|opPgO72yh@@(iq6mesdK0mSqUdL zQ+aqnJs=5+ss0^(9-Y0KT8AnSc?LFD>IA;FULw>D(|PaJVZzM*&H~+ORZ4Q|EI3Yg z4%k<`gjdyAV(k>!RQJX z{<@R1_W}p!Yf;9FEv%FI%!G>fE<&nLUCNP5}raP$f$sg4!th=ZfA+Tg{+t- z2wY11?6(EM&p(~{}NYBLBqQ%?vozgVTc2eR$M zf?FiN>{9eD;2gJmZS-s6v)<`dd9z`Ar!^D(%=U>7-85Tt$Mv#<7H=mmKn_HAtKsgj z6ise-2?n5aQ}mWt)BZYrvARe1h=aVt6n}avu_6Y z7&-%mC1vP2mCc+KBIKz!X-E~*c<&&md+o}z1jz}W zG6jpJ`5wKS*`#AdQB0=)M0}EHa`mn88Qw+^(7;wQN<%`m)AhFMiv>F+o@>1BJzXj;$ z4|xwTR&Yk7ppTPpr%B$)tN^HuhZr>kNzw0pmi5BE_ubCr@c(N-{#=%lV>ORcDWoOU ziCeGAJehBKencntVvR;wyTn~DZfN`&F)vxZ7_gZ1f_tsT>7U3(vI``Aa0SdCEguj$ zt)pXb0t=OoG901Q-=Pn@WQ4_VMw$mCno*h`0rnxuQ6!$q+zT5i*X-Tby{n7h=JU{S zN!ulk;wj{ZKq&B(<2Tj`-jDS(%xtBa#>0h30R053Dm%l1?>3%bt^x)0wZ{6{zvK2^ zp3h}pd^7~m)6`%nu#&}yvv+yP_22SMi@V@!^`0ckTW{{67~b;@=%SmE9%48sxw@9_ z)fCQ2sSn_2Uff=$`nFz8CGR9ZQUkMX`EMYQavN@nBSZCS%t!hM;~Z)0U@$~x)t>el zH*|0P?kFu-fe|2ptIl3P+k>h*+>(3?>)=(vc3`(S6@Q+`y%r>@v;LVv8najT~~O&+FQElB;bl)@D4YhprXcl0oEuZJ%B zMEg}a>r!?GYTA1Dy-g)jd*jnZgCm`MWzat$`yZ^Eyn2A&`zY`UhTN3QDqh;&@O`ab z*y2$&>jd9^!y}N|BV(F0L9_YUTnG~MCbONlOt%58l;CH@o|4gqULsl5Ga9Y0Dy%HJ zpK*#Z+h>U>-;x8EM3{^&QND}?!=jGl*NZAeD2x$6Ctyde=&B1yqrgBwG3#Q)rZ^cG z7!m^D)i~v*$S%J)gxz{`O$N`r?w{u04cReekKs{%;7Dfej3cJaNh?BgjS zSS0NKA9#xG%)fR{TrGwaZ~zZnO?n;~x;_oZQ6K*aU!6J)`de_nnzX7zNcTpK96(rX z5@88fg{Xp4m3fQNWpv-&GmX@B>yK0~4)t0XDcBcD`Wh{pt139axGiOrQhK^L`jc%C zq)vgJtfuS1A(t))ARS5nTE1?q|J!mck)F)dG{qz7uJ>0tb$}I8Vr(%fM7_GE#idc0H&DbO>7ptefzb$_fK)kd_cs8h<_5g+Wt3*vm4`wBR8n<42i73X z&&#`a`d_C5#Ah-Bl^o}9tcG7rvD@Rx@if7O?3r+N@u?WkU{|JXfI;8Ew3yA z8M~Np>I^=)deS!hByuee6~PrM+u{6(1dKx;YCaH*LWUpP$^v-BvTj|+C_hUNYkh3< zwc|*(Ko_Z8aVl8X*Aoshid{a7FOrwzzYCS;PEu_E&^ct_Ir{ZfGi=Kx5j9DWbX|l8 z0Lt3my!pC?d>&(gzWgcsRdRbFhPGjl>kvJ`e zC@Z8n=jyHF9&&(yu8+V22SIC2LWG4FXomnJ{s0oS7e7`;OMhatnlXyTW1LboKEAj$}X=++HFF?ijGfnF$_Ad_Q{EnUg)9V% zEwYt+GSg52ND*!VfTINzAB3ST9otf$(viO;W&jcx)SrY6N0DoI^sE6dg0a#&NoLJCF?rYO3m`n2GN0PzTjbWx+N;?22XMn?K&` zIIM6i-L>&RZ~%n~?Y8v#JoRzv92X+OG7S^B_w5SuQQN(?;9TtgLeYT>d5;x$)7*n8 zA(N+`bPV^S42})DlLHtw#Hr=cG2;!;vhESymXHXC=zh7VBPNtH?|gOHmXQ|j)z<~* z%O-yyFgxIi4C$6jrX>4^17vtuk`um(5t9~5{=*~I>MXzF&r&jV%NJ=N1OPxjKDawU z7K#0V7Vo)m2~x=PGpf@~7$EA>CLzI@hdcoP*d-`Pt2Ng+tJW{G!PhR%vVYC#={mF&g4bV z#c^*${Tg%VI0M6-JONoMZ{93<)#6DZn8Nssw;UQcACqItN<4d zm*M-t>>94{?X1B6K3p61vfeIac$RKB&_hUz%mb&sqR&F%9r`8!FXM&=n-ZL>8PB%M z;cAg3<7$1p48#Ck)xkPUi18NBKmCqf^K3VJG}beX36&sFGAXH4L~WHo2)eDavzay< z+B0$_3h25d%Z@KVOAI&kHiCBm4V!rZedqQC=1WTaiFP!2}nU0ml|M#Ukf!HP+r02Rql{FXpW%b33MF`$hj6o zm;h0-{@^VVl&_J_Y6XCCDB#PSbt1mhVxLo4$GWm8b1123LJm<>=}vMF-hVibK!JM4 zy^**<*c*0(ICXJqjQ@~c%a32Z=+L)#JZJe$@7=l-A#>blL#Ji$YNb-kmc zdB8+S9I?v_(K|@M+I}+F%{hQXqB<1|a`A6Zo<87(TIemguU`t1#C7rhcW{%ZQtB|p zsL`cZa8QJty+ja2RT@RhMlBuCvg1RcY8f1E4w-_R8^`&JNf>BtABMY1j)yLU{usvW>=f-N|2g%t25_)`wTAlL*A;x7HzIh`y!1geWx;=juBY3~fi zfK-Gvd(ub|f4_tSUOEpjR7-a(<4)Z0?L*`a;c^FmV_fNid+&O_bKS#Fj6p~crS9{L zlo2X$1r@lnJ4lK7Qo)4Ei+RbRnEY$6zQC0pks|$L8Qfaf7*8;v>6tKEp=H~kE(qJ- z1haWUNwc7JNaVXWXGn++#eB%+`S`Wj9b1&E9`KZGJ2UOsX;}W5NB90l``UXNF$mKF zWc`g>qr{VP1Yq!b4aVZ_reP8k2=1>!PI9EKm|&3z774{>w$~41&9_QG^q;(OpJ@bg z*XIiXr2DLL4_De)>~Y?^LQ6n*RBfYXuB3cJ04JlH?_!C~OrM$bgb3y&-ag{23Vhtb zzNK{^&S7CZIr{UNvx^qufv7kXGobC8LruyE#PVBfvx!i_>B2dovEvi)>jxan@#AH( zIbdC(Nn&1AeMl?bpPr1jlvrmB!~$DH zkb20TYZR5>%Fg=nKY+>!8>L67pQpi)V;+MjfS>EJ=3;lE^WAphh#Z)t83Z>fbvk@Xo<3%NwPErvq)14*(8AcRTU-njDOxKEn|?Y*6T7 zAKU^gnDEgFPXlr1wa}BzVkRV@efmh4Ip11@5xUs}5dOI|qg~mhG@k|^G{qT&>nS{y z8!CA{GEv-UNUj+`@QRS2c`N6ez`v2LX(b*M5qV}55_&{jWNj2h*TXKH>?p!UxS-ya zr1laLVP}WrmU2EU*{QNw_piuvLzt`%VI2&aEJm5puG>qkT$Fx<|JV8=((B`seDXq9 z#~$zwbxe$Q$lIVu9|u?0>@-zTML%IsUDBV}uFZ~{!iuNQ0#f#Ol0G%79)aI#oTScN z2=WY*=A|CNedI?cu5p)r;UMW33YhCorm(T>*UaB>)%g6GQ_Fe2S5E(ip^xYA`We{`R*ywU5Qk0%hz&tA_f!sq9gh&0YtmaoTaxhu}8@vu%5k| zAhQm*t0Pv*ibDkn{lW$-F?vx8vOmI~0xZLz5N^a-aFG$$udB9qbV6iUkVx&t|!Jn zKquX6!FrL7p8cNVzTWF!mWcfeOe;Ty7;noN33PY?=}?aW1;cl7H@Y8l#WoZY(ybU% zd%e9@msMBQdb*w3jroA*n9OWTx`>HvYpstw!Ws#kbwq|V3|gceT@ExclKW;_Hs9#$ z#p$vG86oMt973L>;QMH|U$Oq62{W`@bxdG&lRa?UD4#C_qEf;Xm3sbojP=&f3JE^I z*ASUnXGDVB#3}F|JbCpK?g%QB#8<9cWs%2bK1#r!zce4Ev)KTN4)!eOPYyPS?;fB5 zQ8BNQ5O)_+yfL zQ72cx>(0c1OW>5v8fUf3zwLwr#5ej(67_8|Bw1@uxWAP6&cW6v5!c zdhac`I)o}skh&Z@1=1r(ao~>e z#92M4MFo>-qK`r{VvAG$uNg0BMyJQCmC_q?-;fD3q^dX*^Y$PsZv6|JdMujeQx@K z<-dLma8M*VUzj2UNPD06Dz}F6fPCU?9Ju?`#=>zFV)5L6ctJA>ZSv)(dTyiu7`a9- zdyWi2_^eS#v<{ErE*u1slfh^mtIC#P{g>-xu{kGSk_HCMLG%_&5v=2gCe}Jy+?8?l zyCUd^6Z9h=f|4+R#*gtnmpc?J#6GnP;`QcE7R@X%aRtWNeG6^!ut%|K2WQ|g&}w3% z>?cOVs*SjUs=k1%&y89ut0h2DBiMn@nulO2f3Eg+0rfyaP?;rXODBm>s1SV^KRRSq zy4nD%o0sQ6fsSPskfFUP4{U+eDm)0zJH-v{YtWb{*I(`sL18?lHqdfZUfq9LO7FJ_ z^c9Br|A&v|dJpP;k54sx5L^;tC}q#CcZi$=7jhpdAAGiu{OhA?W;26nJx%2Y)c>9n=9C}M*9SWo9qTIYp3@LwSbx; zHr^b2PN;R)2WRrmAib%QaRyW6TRvm9B$aeNKwtV%-FkU+K3_M3J?STg7fg5;9^ zZy(?E-jl%+3i}z5pyBy|ou|{t_e(N43y$gdWHuMT%nY)RgbZcQZzL7LpbiG(jku7A z%<5^$7&|D;&sYfYS+5V5szIgfzg7P+(qo}uXLu-IhnZ9t!EO4Lz90ANR2HaFgqtS=+yA0sWW*HgXY24QLgNoY6 zkpuoP_7X>0?SQ{;!5=AXQe&F&kPONvv>!_OZ5!WkuL44Z^F}xV#$r?9fDNVi(tl@A!6RU1K(JoWm0)gklYq8TgfuyCXe2z()7y+7l#)V z_S3Wq?1r!ddz4{o$%5WR$}LDwYXRz7=*w~-mh*fd_^1v)1UGWPu9KJsKLdeqBH0GI z`wa7I3NYEv+wIL%TJmJ_nq+UN#T++85?bqzc3F3}KP1btg&y-zih7+Wwl`y4U5^iK zeSBx@E%-XXabrRK$E-ZlD~)+ao43>Q;(?AuKBQBsOtbIb{^;VIt7OA@-rCkg0TKXn zr|}5f&pXC3d%lSmA!Piu(fPB=)z*jP2p`gq$!L!D(x!qh{uaFh|Hnp7x%OBE1nhi(B4)#y~kFpN|ZZ^dF%)!}0;)GuN$AsW2fMs<^SAL4R_|-Y`+62m1{BI=InXVwN z_gwqSlZyDa4#E=^7sr-(0S%Z72Dv=k8Z6yHVs+@PMoA-FO7^7Vfx=&EN!JU{7A4kn z>ApOEXZYXuZ$Ca!9Zx-JDR(nJv2>oyLfb^V^si{2%wBe@SxB%Kl$}bS*95mD*DGFVU4aHdU~U3M3TXwmg5# zj}dXi{L04=-A1yFQ$X|w#yhPy-2$&-ncmAblN8o)nx|(GDP9W9c5&ektUWR*Wb^!V zn#UvQdNMW>o$ zSv{u28(Dp8Zofo_H!_4f@H@VbG9Dq#mbSdl)zy#=C5Z-?RNOb;?8~K>WMdqXVPv(6 z!`tpU5n_ygFiO<}U5esWl0FA^aM>cXP~rXUw8K+D>-vWVhSF^5=4j~ z{`}=pY9B9VdUbKXd%Hoa4}9v^#aZ<+N;v+ADEbnjGY|78;){i|W{F_G@UrKpbe!8| zzO8S7(G7_ciIc1#=_T4K3QHvV2aRFvYrUrMi#TSM5?jA#jK~t6+(0}6j2Li3@JIOn z^~de5D(*yLOADkk_~pBvxyTD7E=g zulIe*K^xChkcb~}{9x4##rB*^SO5K4B*;nYJLG1~jXRY-a=4m$yX1RK!qcw_mv^dE zu57T^C3}OKFCKd#&r&C!TDAEhIqAC*xnJAf{K#C-W8(sBXpBA`cb|EttK*&Cp2T4# z^B1SMwe47E2Yj>fcem+l{v~s~yKN`WBNpC05B{(a>?(8>!-Ti!5<2Z_xNF&<+q*;v z>HQl?+_X|)@obq>q%NcTGB{EYu44D&uW&jpKA2ejni7l0N~C50_kmlK*Zzm?xkp2E zzUFfa-(Rd&{);nt6jIJ-{1eq{sO&&%6(0$56aow!!IO?irjfz<6cyg2R}#J|95*M% z083acSvOdqpWMyLnzMV4*Vn5%H-!JflCA&ilOH?vb{tS7?y( zo%jkj|aSyo= zwd)odxBX-$#hvw2gp?XxMezx8wWWi}Q` zUGOVC-YChy8fUAkrRlu(qu+Deuid7e)*Z#gGp-^*jBA=^oLLH`HuUdF&;x1%ICgfc znZIb7xc_W3Tn%*ektXXyfqrf8{P~#T4@<}0`?K3JR=;<%<-D^skjL9{wH29tff~6c zHHjzwbEodA-J-%+<0e+q#pIsM_UW8++KFjr=VI0So&!+jD>EA#+(l%ZuIdv`zbr_bg9#l$}+GzYUXF(Z$LB$L% lU>N4Tg!ca=RsHX>bGp$Q6HI)cr