Successfully reported this slideshow.
Upcoming SlideShare
×

# Business statistics final

463 views

Published on

Published in: Business, Technology
• Full Name
Comment goes here.

Are you sure you want to Yes No
Your message goes here
• Be the first to comment

### Business statistics final

1. 1. 2012. 12. 10 Business Statistics - Disaggregation of energy saving action - 1
2. 2. 0. Outline 1. 2. 3. 4. 5. Objectives Data Analysis process Result Conclusion 2
3. 3. 1. Objectives  Background  Highly interested in human behavior concerning energy saving since the disaster on 311, 2011 in Japan.  Data analysis on amount and actions for energy saving is important to understand this condition.  Problems  It is not still unclear how the relationship between amount of energy saving and energy saving actions was explained.  It is not still unclear how much energy saving is and which actions can contribute to energy saving. 3
4. 4. 1. Objectives  Objectives  To understand the energy saving actions  To disaggregate the amount of energy saving by each action through multi-regression model Energy Saving Action Owned Applian ces Easy to act 4 Amount of Energy Saving Temp rature Time Spendi ng at home High effect
5. 5. 2. Data  Overview of questionnaire  The questionnaire investigation conducted for 20’s to 60’s on the web Period Conducted in May. 2012 Area Around Tokyo (Tokyo, Kanagawa, Chiba, Saitama) Screening condition 1) No moved since Dec. 2010 2) No changed family structure since Dec. 2010 3) Sample having electric and gas meter receipts on Dec. to Mar. 2010 and 2011 # of samples 5,892 ss 【Age - Sex sample rate】 20s 30s 40s 50s 60s Male 1,100 1,094 813 1,176 Female 5 274 248 296 302 238 351
6. 6. 2. Data  Making data To sum up the amount of electricity • To extract from meter receipt of electricity from Dec. to Mar. on 2010 and 2011 To calculate the amount of energy saving • To calculate the amount of energy saving on 2011 winter compared to 2010 based on the corrected amount of electricity 6 To correct for the influence of the date of metering To correct for the influence of temperature difference The effect of each actions for energy saving • To conduct multi regression analysis with the amount of energy saving as objective variable and each action as explanatory variables
7. 7. 2. Data  To correct for the influence of the date of metering  Corrected the amount of electricity because inputted data by sample through questionnaire is based on each date of metering  For instance, （The amount of ele. on Jan.）＝ （date of metering）×（The amount of ele. on Jan. / 31）＋ （31-date of metering）×（The amount of ele. on Dec. / 28） Dec. of metering Date Jan. of metering Date Meter receipt on Jan. Feb. of metering Date Meter receipt on Feb. Real amount of electricity on Jan. 7
8. 8. 2. Data  To correct for the influence of temperature difference The amount of electricity [kWh/month/house]  To make single regression model using the amount of electricity per a house on Dec., Jan. and Feb. from 1998 to 2010 in TEPCO area and temperature in Tokyo.  To definite this coefficient as a temperature corrected coefficient (decrease electricity of 11.1kWh(Dec.), 11.5kWh(Jan.) and 9.0kWh(Feb.) per 1 degree Celsius.)  Decreased electricity to correct -2.5 degree C(Dec.), -0.5 degree C(Jan.) and -1.7400.0 degree C(Feb.) on 2011 compared to 2010. 390.0 380.0 y = -11.11x + 438.1 R² = 0.641 370.0 360.0 y = -11.53x + 439.5 R² = 0.530 350.0 1月 Jan. 340.0 330.0 320.0 2月 Feb. y = -8.980x + 387.7 R² = 0.455 310.0 300.0 4.0 8 12月 Dec. 5.0 6.0 7.0 8.0 9.0 10.0 Temperature [degree C]
9. 9. 3. Analysis process  To know deeply the objective data and find the correlation with various data 9 1. Overview ing the objectiv e data 2. Making the correlatio n matrix 3. Picking up explanatory variables 4. Developing the multi regression model 5. Improving the multi regression model
10. 10. 4-1. Overviewing the objective data  Average energy saving  Calculated the amount (82.9kWh down) and the ratio (7.1% down) of energy saving by using the corrected electricity by date of metering and temperature from Dec. to Feb. on 2011 Dec. Electricity Jan. Dec. Dec. to Feb. 2010 368 kWh 444 kWh 364 kWh 1,175 kWh 2011 333 kWh 415 kWh 343 kWh 1,092 kWh 34.4 28.2 20.3 82.9 kWh kWh kWh kWh 9.4 6.4 5.6 7.1 % % % % The amount of energy saving [kWh] The ratio of energy saving [%] 10
11. 11. 4-1. Overviewing the objective data  The overview of the amount of electricity saving  The amount of electricity conservation is normal distributed with the center of 82.9kWh on average  70% of sample could accomplish energy saving 70% 11
12. 12. 4-2. Making the correlation matrix  The definition of explanatory variables  Heard the degree of 3 segments of energy saving actions based on questionnaire survey  Explanatory variables are defined by the difference of the degree of all the actions between 2010 and 2011 Saving Action The degree of electricity saving actions based on 7point scale Detail setting temperature and using hour of heaters 12 Owned Appliances Dispose, purcha se and replacement of each appliance Hour at Home Hour with nobody at home each weekday and holiday
13. 13. 4-2. Making the correlation matrix  The overview of explanatory variables  Calculated the increasing ratio of energy saving actions on 2011 compared to 2010  Some results are shown here, totally all of energy saving actions were increased than 2010 Close the door when heaters are working 13 Unplug the appliances Set refrigerator low level Boil water by gas cooking stove Stop rice cooker to keep warm
14. 14. 4-2. Making the correlation matrix  The overview of explanatory variables  Investigated the time of use and set of temperature for appliances, especially heaters and lights, on 2011 compared to 2010  Some results are shown here, totally all of appliances were not used a lot Lengthen 14 Shorten Lengthen Shorten
15. 15. 4-2. Making the correlation matrix The amount of electricity saving The amount of electricity saving Time of Light in Living room Time of AC Unplug the appliances Time of Light in Bed room … 1 0.21 0.15 0.15 0.13 … Time of AC 0.21 1 0.17 0.11 0.12 Time of Light in Living room 0.15 0.17 1 0.14 0.16 … Unplug the appliances 0.15 0.11 0.14 1 0.07 … Time of Light in Bed room 0.13 0.12 0.16 … … … 0.07 … 1 … … … To find the explanatory variables which have the strong relationship with the amount of electricity saving. To categorize the similar explanatory variables not to include multicollinearity. 15
16. 16. 4-3. Picking up explanatory variables  Top variables which have strong relationship with The amount of electricity saving Time of AC -0.208 Set refrigerator low level -0.145 Close the door when Unplug the heaters are working appliances -0.149 Boil water by gas cooking stove -0.140 -0.153 Stop rice cooker from keeping warm -0.146 Time of Light Time of Light in Bed Time of TV in Living room room -0.153 -0.131 -0.125 16
17. 17. 4-4. Developing the multi regression model  Based on hypothesis and statistical approach, I developed the multi regression model.  Hypothesis is the most important because model must be easy to explain and be accepted to audience.  Then I tried to find the optimal explanatory variables without decreasing p-value, AIC and R^2 Hypothesis Statistics A variable E variable B variable C variable D variable 17 Objective variable F variable G variable . . . .
18. 18. 4-5. Improving the multi regression model  Step up explanatory variables  Step up explanatory variables from the fundamental factor influenced to electricity  Made three models by adding variables in turn with checking AIC and p-value Electricity saving Purchasing appliances Model 1 Set temperature and Used hours The degree of use for appliances Used hours of TV and lights Model 2 Electricity saving actions Model 3 18 Hours at home
19. 19. 5. Result  Result of model 1  Model 1 is the simple model based on variables showing the purchase of appliances Coefficient P-value (Intercept) -59.858 P<0001 Oil stove -61.884 P<0.001 Electric stove 27.921 P<0.05 Gas fan heater -97.654 P<0.001 LED -42.272 P<0.001 Television -25.383 P<0.001 Humidifier 29.148 P<0.01 -33.323 P<0.05 39.872 p<0.05 Refrigerator -43.375 P<0.001 Dish washing machine -94.446 P<0.01 Hot watering toilet seat Washing and drying machine AIC = 73,886 19
20. 20. 5. Result  Result of model 2  Model 2 has used hours and set temperature of appliances Purchasing appliances Coefficient Pvalue (Intercept) -52.227 Oil stove -36.630 P<0.001 Electric stove 25.229 P<0001 P<0.05 Used hours AC Electric carpet Pvalue -16.123 P<0.001 -7.581 P<0.001 Gas stove 9.405 P<0.1 4.429 P<0.1 Gas fan heater -75.162 P<0.001 Gas fan heater LED -36.574 P<0.001 Oil stove Television -20.834 P<0.001 Oil fan heater Humidifier Coefficient 13.932 P<0.001 4.743 P<0.1 28.834 Hot watering toilet seat Washing&drying machine P<0.01 Electric stove -12.388 P<0.001 -37.420 P<0.05 Halogen heater -10.119 36.334 p<0.05 Electric fan heater -17.945 p<0.001 P<0.01 Refrigerator -45.922 P<0.001 Oil heater -21.057 P<0.001 Dish washing machine -81.189 Lights in living room -15.646 P<0.001 Lights in bed room -17.005 P<0.001 TV in living ronnm -8.956 P<0.001 Water server 20 P<0.01 43.134= 73,377 P<0.05 AIC
21. 21. 5. Result  Result of model 3  Model 3 involves energy saving actions Purchasing appliances Coefficient Pvalue Used hours Coefficient Pvalue (Intercept) -42.499 P<0001 AC Oil stove -30.476 P<0.01 Electric carpet -6.009 P<0.001 26.950 P<0.05 Gas stove 11.628 P<0.05 6.751 P<0.01 Electric stove -13.662 P<0.001 Gas fan heater -65.947 P<0.001 Gas fan heater LED -27.793 P<0.001 Oil stove Television -15.110 P<0.05 Oil fan heater 6.726 P<0.01 Humidifier 20.130 P<0.05 Electric stove -10.542 P<0.01 -38.548 P<0.01 Halogen heater -8.973 P<0.05 31.711 p<0.05 Electric fan heater -15.898 p<0.01 Hot watering toilet seat Washing&drying machine 13.870 P<0.001 Refrigerator -49.654 P<0.001 Oil heater -19.404 P<0.001 Dish washing machine -92.589 P<0.01 Lights in living room -11.019 P<0.001 49.583 P<0.05 Lights in bed room -13.907 P<0.001 Water server 21 TV in living ronnm -5.084 P<0.05
22. 22. 5. Result  Result of model 3  Model 3 involves energy saving actions Energy saving actions Set low temperature of electric floor heater Coefficient P-value -132.945 P<0001 -13.962 P<0.01 -3.267 P<0.1 Close the door when heaters are working -20.267 P<0.001 Unplug appliances -12.435 P<0.05 Set refrigerator low level -16.082 P<0.01 Boil water by gas cooking stove -16.636 P<0.01 Stop rice cooker from keeping warm -20.869 P<0.001 31.711 p<0.05 Used degree of Hot watering toilet seat -49.654 P<0.001 Used degree of washing & drying machine -92.589 P<0.01 49.583 P<0.05 Set low temperature of gas fan heater Hours at home Used degree of humidifier Used degree of electric pot AIC = 73,190 22
23. 23. 5. Result  The improvement of models  The improvement of R^2 is shown  Still low but can extract the effective energy saving actions which have high t-value  It means that results show the disaggregated electricity saving amount by effective and important actions Model 3 Model 1 Adjusted R^2 = 0.0358 23 Model 2 Adjusted R^2 = 0.1185 Adjusted R^2 = 0.1493
24. 24. 5. Result  The residual analysis  The big difference between active and passive energy savers is whether they purchased new appliances  Positive energy saver  Tends to purchase much new appliances such as LED, TV and oil stove  Passive energy saver  Tends to purchase much new appliances such as fumidifier, electric stove and water server Passive +3σ -3σ Positive 24
25. 25. 6. Conclusion  The feature of effective energy saving actions  Summarize the feature of effective energy saving actions from the result of Model 3 • The effect of purchase of new appliances, especially electric heater, LED, is highest • Reducing used hours of appliances is more effective for saving energy than setting low temperature • Switching to gas and oil heaters contributes to saving energy due to the avoidance of electric heaters including AC • Reducing the use of electric heat generator such as humidifier, Hot watering toilet seat, electric pot, drying machine and rice cooker is definitely important to save energy 25