Possible Problems of Local Finance Fiscal spillovers: When taxes based on property values, people in low valued houses can free ride the median voter chooses a level of school spending that is lower than if the district was homogenous Answer: Can be solved using zoning regulation Results in human capital segregation Centralized finance leads to human capital integrated schools Answer: People will sort themselves Tiebout style if there is high correlation between
View full slide show




VOMmean on WINE150hl(FA,FS,TS,AL) 0 1 1 4 2 4 3 5 4 3 5 8 6 7 7 1 8 7 _________[0 9 2 . 10) 42 low 0 high CLUS_1.1.1.1.1.1 _________[0 . 19) 56 low 10 5 [10,16) 15 low 12 high CLUS 1.1.1.1.1.2 [19,22) 1 low 12 high 11 4 But no algorithm would pick 19 as a cut! ___12 _____[0 . 13) 56 low 5 1 low 12 high But no alg would pick a 13 cut! 13 7 [13,16) 14 3 _________[0 15 3 . 16) 57 low 12 high CLUS 1.1.1.1.1 35 high CLUS 1.1.1.1.2 17 2 [16,31) 0 low 18 5 19 4 20 3 21 1 22 4 23 4 24 5 25 1 26 2 27 1 29 2 30 1 31 1 _________[0 . 31) 57 low 47 high CLUS_1.1.1.1.1 32 1 28 high CLUS 1.1.1.1.2 34 1 [31,58) 0 low 36 1 37 3 38 1 39 2 40 1 43 6 44 4 46 2 47 1 48 1 50 1 52 1 _________[0 .58) 57 low 75 high CLUS_1.1.1.1 56 1 5 high CLUS 1.1.1.2 60 1 [58,70) 0 low 63 1 65 2 _________[0 .70) 57 low 80 high CLUS_1.1.1 67 1 [70,78) 0 low 2 high CLUS 1.1.2 72 1 _________[0 .78) 57 low 82 high CLUS_1.1 74 1 82 1 7 high CLUS1.2 83 1 [78,94) 0 low 85 1 86 1 87 2 _________[0.94) 57 low 89 high CLUS_1 88 1 99 1 0 low 4 high CLUS_2 105 1 113 1 119 1 d=VOMMEAN DPP on WINE_150_HL_(FA,FSO2,TSO2,ALCOHOL). Some agglomeration required: CLUS1.1.1.1.1.1 is LOW_Quality F[0,10], else HIGH Quality F[13,119] with 15 LOW error. Classification accuracy = 90% (if it had been cut 13, 99.3% accuracy!) 7 1 8 4 STDs=(1.9,9,23,1.2) 9 4 maxSTD=23 for 10 5 11 4 d=e TS on WN150hl(FA,FS,TS,AL 12 7 13 7 14 8 _________[0 CLUS 1.1.1.1.1.1 15 2 . 16) 42 low 12 high CLUS 1.1.1.1.1.2 16 5 [16,22) 15 low 17 4 18 5 . 19 7 20 3 _________[0 . 22) 57 low 12 high CLUS_1.1.1.1.1 21 3 32 high CLUS 1.1.1.1.2 23 2 [22,33) 0 low 24 9 25 3 26 1 27 4 28 4 29 5 30 1 31 2 _________[0 . 33) 57 low 44 high CLUS_1.1.1.1.1 32 1 31 high CLUS 1.1.1.1.2 34 2 [33,60) 0 low 35 1 36 1 37 1 39 1 41 1 42 3 43 1 44 2 45 1 47 6 48 4 49 2 50 1 51 1 53 1 55 1 _________[0 .60) 57 low 75 high CLUS_1.1.1.1 59 1 5 high CLUS 1.1.1.2 63 1 [60,72) 0 low 65 1 67 2 _________[0.72) 57 low 89 high CLUS1.1.1 69 1 74 1 [72,80] high CLUS1.1 CLUS1.1.2 _________[0.80) 57 low 892 high 75 1 84 1 7 high CLUS1.2 85 1 [80,95] 86 1 87 1 88 2 _________[0.95) 57 low 89 high CLUS1 89 1 100 1 4 high CLUS2 106 1 113 1 119 1 Identical cuts and accuracy! Tells us that d=eTotal_SO2 is responsible for all separation. WINE
View full slide show




VOMmean w F=(DPP-MN)/4 Concrete4150(C, W, FA, Ag) 0 1 1 1 5 1 6 1 7 1 8 4 med=14 9 1 10 1 11 2 12 1 13 5 14 1 15 3 med=18 16 3 17 4 18 1 19 3 20 9 21 4 22 3 23 7 24 2 med=40 25 4 26 8 27 7 28 7 med=56 29 10 30 3 31 1 32 3 33 6 med=61 34 4 35 5 37 2 38 2 40 1 42 3 43 1 44 1 45 1 46 4 ______ CLUS 4 gap=7 49 1 56 1 [52,74) 0L 7M 0H CLUS_3 58 1 61 1 65 1 66 1 69 1 ______ gap=6 71 1 77 1 [74,90) 0L 4M 0H CLUS_2 80 1 83 1 ________ gap=14 86 1[0.90) 43L 46 M 55H 100 1 [90,113) 0L 6M 0H CLUS_1 103 1 105 1 108 2 112 1 _____________At this level, FinalClus1={17M} 0 errors C1 C2 C3 C4 med=10 med=9 med=17 med=21 med=23 med=34 med=33 med=57 med=62 med=71 med=71 med=86 CLUS 4 (F=(DPP-MN)/2, Fgap2 _______ 0L 0M 3H CLUS 4.4.1 gap=7 0 3 =0 0L 0M 4H CLUS 4.4.2 gap=2 7 4 =7 9 1 [8,14] 1L 5M 22H CLUS 4.4.3 1L+5M err H 10 12 11 8 gap=3 12 7 ______ 0L 0M 4H CLUS 4.3.1 gap=3 15 4 =15 18 10 0L 0M 10H CLUS 4.3.2 gap=3 21 3 =18 22 7 ______ 23 2 [20,24) 0L 10M 2H CLUS 4.7.2 gap=2 25 2 [24,30) 10L 0M 0H CLUS_4.7.1 26 3 27 1 28 2 gap=2 29 1 31 3 CLUS 4.2.1 gap=2 32 1 [30,33] 0L 4M 0H Avg=32.3 34 2 0L 2M 0H CLUS 4.2.2 gap=6 40 4 =34 ______ 0L 4M 0H CLUS_4.2.3 gap=7 47 3 =40 52 1 0L 3M 0H CLUS_4.2.4 gap=5 53 3 =47 54 3 55 4 56 2 57 3 ______ gap=2 58 1 [50,59) 12L 1M 4H CLUS 4.8.1 L60 2 8L 0M 0H CLUS_4.8.2 61 2 [59,63) gap=2 62 4 ______ =64 2L 0M 2H CLUS 4.6.1 gap=3 64 4 [66,70) 10L 0M 0H CLUS 4.6.2 67 2 gap=3 68 1 71 7 ______ gap=7 72 3 [70,79) 10L 0M 0H CLUS_4.5 79 5 5L 0M 0H CLUS_4.1.1 gap=6 85 1 =79 87 2 [74,90) 2L 0M 1H CLUS_4.1 1 Merr in L Median=0 Avg=0 Median=7 Avg=7 Median=11 Avg=10.7 Median=15 Avg=15 Median=18 Avg=18 Median=22 Avg=22 2H errs in L Median=26 Avg=26 Median=31 Median=34 Avg=34 Median=40 Avg=40 Median=47 Avt=47 Accuracy=90% Median=55 Avg=55 1M+4H errs in Median=61.5 Avg=61.3 Median=64 Avg=64 2 H errs in L Median=67 Avg=67.3 Median=71 Avg=71.7 Median=79 Avg=79 Median=87 Avg=86.3 Suppose we know (or want) 3 clusters, Low, Medium and High Strength. Then we find Suppose we know that we want 3 strength clusters, Low, Medium and High. We can use an antichain that gives us exactly 3 subclusters two ways, one show in brown and the other in purple Which would we choose? The brown seems to give slightly more uniform subcluster sizes. Brown error count: Low (bottom) 11, Medium (middle) 0, High (top) 26, so 96/133=72% accurate. The Purple error count: Low 2, Medium 22, High 35, so 74/133=56% accurate. What about agglomerating using single link agglomeration (minimum pairwise distance? Agglomerate (build dendogram) by iteratively gluing together clusters with min Median separation. Should I have normalize the rounds? Should I have used the same Fdivisor and made sure the range of values was the same in 2nd round as it was in the 1st round (on CLUS 4)? Can I normalize after the fact, I by multiplying 1st round values by 100/88=1.76? Agglomerate the 1st round clusters and then independently agglomerate 2nd round clusters? CONCRETE
View full slide show




More Sort Names • Bubble sort · Cocktail sort · Odd-even sort · Comb sort · Gnome sort · Quicksort · Selection sort · Heapsort · Smoothsort · Cartesian tree sort · Tournament sort · Cycle sort · Insertion sort · Shell sort · Tree sort · Library sort · Patience sorting · Monkey-puzzle sort · Merge sort · Polyphase merge sort · Strand sort · American flag sort · Bead sort · Bucket sort · Burstsort · Counting sort · Pigeonhole sort · Proxmap sort · Radix sort · Flashsort · Bitonic sorter · Batcher odd-even mergesort · Timsort · Introsort · Spreadsort · UnShuffle sort · JSort · Spaghetti sort · Pancake sort
View full slide show




Motor Controls Actions Left Forward Left Reverse Left PWM Left Motion Low Low Low Low High High High Low Low High High Low Low High Low High Low High Low High Low High High High Coasting Coasting Coasting Reverse Coasting Forward Coasting Active Braking Right Forward Right Reverse Right PWM Low Low Low Low High High High Low Low High High Low Low High Low High Low High Low High Low High High High Right Motion Coasting Coasting Coasting Reverse Coasting Forward Coasting Active Braking 13
View full slide show




VOMmean w F=(DPP-MN)/4 Concrete4150(C, W, FA, Ag) 0 1 1 1 5 1 6 1 7 1 _________[0,9) 8 4 9 1 [9,14) 10 1 11 2 12 1 13 5 14 1 [14,18) 15 3 16 3 17 4 18 1 [18,23) 19 3 20 9 21 4 22 3 23 7 24 2 [23,31) 25 4 26 8 27 7 28 7 29 10 30 3 31 1 [31,36) 32 3 33 6 34 4 35 5 37 2 [36,39) 38 2 40 1 [39,52) 42 3 43 1 44 1 45 1 46 4 ______ 49 1 56 1 [52,90) 58 1 61 1 65 1 66 1 69 1 71 1 77 1 80 1 83 1 _________[0.90) 86 1 100 1 [90,113) 103 1 105 1 108 2 112 1 e4 accuracy rate = 104/150= 69% 2 Low 1 Low 5 Medium 3 Medium 5 High 5 High CLUS_1.1.1.1.1.1.1.1.1 CLUS_1.1.1.1.1.1.1.1.2 5 Low 2 Medium 4 High 6 Low 1 Medium 13 High CLUS_1.1.1.1.1.1.2 25 Low 4 Medium 19 High CLUS_1.1.1.1.1.2 3 Low 7 Medium 11 High CLUS_1.1.1.1.2 1 Low 1 Medium 2 High 0 Low 12 Medium 1 High 0 Low 11 Medium CLUS_1.1.1.1.1.1.1.2 . 0 High CLUS_1.1.1.2 CLUS_1.1.2 CLUS_1.2 43 Low 46 Medium 55 High CLUS_1 0 Low 6 Medium 0 High CLUS_2 d=e4 Conc4150 ________=0 0 17 ________=1 1 11 ________=3 3 12 ________=6 6 35 _______ ________=13 13 25 _______ 22 25 =22 24 8 =24 44 7 ________=44 67 ________=67 4 89 2 ________=89 91 4 13 Lo 2 Lo 12 Lo 13 Lo 3 Lo 0 Lo 0 Lo 0 Lo 0 Lo 0 Lo 4 Med 9 Med 0 Med 5 Med 3 Med 6 Med 8 Med 7 Med 4 Med 4 Med 0 4 STD=(101,28,99,81) d=e1 Conc4150 6 3 7 2 8 Low 7 Medium 0 High CLUS_1.1.1.1.1.2 12 10 [9,16) 13 2 14 3 CLUS_1.1.1.1.2 18 9 [16,31) 18 Low 11 Medium 0 High 20 4 22 5 23 3 24 5 27 3 1 Low 3 Medium 3 High CLUS_1.1.1.2 31 3 [31,39) . 36 4 5 Low 5 Medium 7 High CLUS_1.1.2 41 2 [39,52) 42 1 43 4 44 3 46 3 48 2 ______ 49 2 55 13 4 Low 17 Medium 38High CLUS_1.2 58 8 [52,80) 60 6 Entirely inconclusive using e ! 62 5 1 65 4 71 16 72 4 ________[0.80) 43Low 46Medium 55High CLUS_1 74 3 7 High CLUS_2 82 4 [80,101) 0 Low 7 Medium 83 7 97 2 100 1 VOM MN 0 Hi 0 Hi 0 Hi 17 Hi 19 Hi 19 Hi 0 Hi 0 Hi 0 Hi 0 Hi CLUS_11 CLUS_10 CLUS_9 CLUS_8 CLUS_7 . CLUS_6 . CLUS_5 CLUS_3 CLUS_1 CLUS_2 d=e3 Conc4150 0 15 3 2 ________ 5 4 17 3 19 8 21 1 29 3 ________ 41 28 46 3 47 8 48 3 52 4 ________ 53 15 58 3 62 4 63 4 64 1 65 7 67 3 69 4 72 3 73 12 75 2 78 5 83 1 100 4 2,4 [0.9) [9,32) 3 Low 1 Low 16 Medium 8 Medium 2 High CLUS_1.1 6 High CLUS_1.2 [0.32) 4 Low 24 Medium 8 High CLUS_1 [32,101) 39 Low 28 Medium 47 High CLUS_2 [32,55) 21 Low 12 Medium 28 High CLUS_2.1 [55,101) 1 Low 8 Medium 6 High CLUS_2.2 0 2 3 4 ________[0,5) 3 Lo 1 Med 2 Hi CLUS_2 8 4 ________=8 0 Lo 1 Med 3 Hi CLUS_3 10 2 ________=10 0 Lo 2 Med 0 Hi CLUS_4 12 8 ________[10,14) 1 Lo 4 Med 9 Hi CLUS_5 13 4 15 4 16 14 17 3 18 1 19 3 ________[14,21) 9 Lo 9 Med 15 Hi CLUS_6 20 8 22 15 ________[21,25) 3 Lo 1 Med 15 Hi CLUS_7 24 4 ________[25,28) 5 Lo 1 Med 3 Hi CLUS_8 27 9 29 3 e2 accuracy= 30 6 31 3 Inconclusive on e2! 93/150=62% 32 3 33 7 34 4 d=e2 ________[28,36) 14 Lo 12 Med 8 Hi CLUS_9 35 8 0 Hi CLUS_10 5 [36.40) 7 Lo 1 Med Conc4150 37 38 3 0 1 2 3 4 5 6 7 8 9 10 11 12 13 18 19 20 21 29 30 31 32 34 35 36 62 64 93 121 125 2 5 6 12 2 1 6 6 1 4 12 11 5 3 2 9 10 4 1 4 9 4 9 4 1 2 5 4 2 4 2,4 =0 2L 0M 0H C14 =1 1L 4M 0H C13 [2,4) 11L 6M 1H C12 =4 0L 2M 0H C11 [5,9) 14L 0M 0H C10 [9,11) 3L 0 M 13H C9 =11 =12 =13 [15,25) 5L 5L 0L 2L 2 M 4H 0 M 0H 3 M 0H 4 M 19H C8 C7 C6 C5 [25,33) 0L 0 M 18H C4 [33,50) 0L 14 M 0H C3 [50,93) 0L 7 M 0H C2 [93,m) 0L 10M 0H C1 Accuracy= 127/150=85% CONCRETE
View full slide show




Centralized v. Local Finance How does the government finance education? Local finance  Taxes collected from a certain district  Used to finance local public good  Tax rate determined by the district’s median voter Centralized finance  Taxes imposed on statewide or nationwide tax base  Redistribution across states: A district gets a share of the tax money, not necessarily equal to money collected from that district  Tax rate determined by the state’s median voter  A district has little control over tax rates or provision of public good
View full slide show




Credit Student Origins High School Graduates GCC Does Regular Outreach to the Following High Schools: Hoover High School Glendale High School La Canada High School Franklin High School Eagle Rock High School Daily High School Crescenta Valley High School Clark Magnet Burroughs High School Burbank Adult School Burbank High School Belmont High School Bell-Jeff High School Los Angeles High School Marshall High School North Hollywood High School Verdugo Hills High School Downtown Magnet Miguel Contreras High School John Francis Poly Taft High School Temple City High School Garfield Campus
View full slide show