Computer Lab #3 Apr 3rd, 2009Analyze data: T-test, ANOVA and CorrelationTips to get the software and data work:To use STATA on Linux systemtype "add stata" in the terminaltype “xstata” in the terminalTo use flash drive on Linux systemtype "add consult" in the terminal type "tellme root" and pay attention to the password it gives you type "attach-usb" and then enter that passwordThe path will be "/mnt/usb/foldername"type "detach-usb", and give the same password to detach f-driveMetadata of “Hedonic.dta”This data set contains observations on house prices and attributes in the city of Newton.id house codeprice sale pricelot lot sizestyle building styleyear_b year when the house was builtsize total areas of living spaceroom number of roomsbed number of bedroomsbath number of bathroomsq1 interior condition of the house: “above”,”average”,”bellow”q2 bathroom condition: “above”,”average”,”bellow”year_s year of saleold dummy variable = 1 if the house was built before 1930STATA commands used in today’s class ttest compare the sample means or other descriptive statistics valuesoneway one-way analysis of varianceanova analysis of variancecorr simple correlation among variablestwoway scatter produce scatter plot of outcome vs. predictorgraph matrix produce multiple twoway scatter plot at a timeMIT OpenCourseWarehttp://ocw.mit.edu 11.220 Quantitative Reasoning & Statistical Methods for Planners I Spring 2009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.1 11.220 Quantitative Reasoning and Statistical Methods for Planning Computer Lab #3 Apr 3rd, 2009 Analyze data: T-test, ANOVA and Correlation Tips to get the software and data work: To use STATA on Linux system type "add stata" in the terminal type “xstata” in the terminal To use flash drive on Linux system type "add consult" in the terminal type "tellme root" and pay attention to the password it gives you type "attach-usb" and then enter that password The path will be "/mnt/usb/foldername" type "detach-usb", and give the same password to detach f-drive Metadata of “Hedonic.dta” This data set contains observations on house prices and attributes in the city of Newton. id house code price sale price lot lot size style building style year_b year when the house was built size total areas of living space room number of rooms bed number of bedrooms bath number of bathrooms q1 interior condition of the house: “above”,”average”,”bellow” q2 bathroom condition: “above”,”average”,”bellow” year_s year of sale old dummy variable = 1 if the house was built before 1930 STATA commands used in today’s class ttest compare the sample means or other descriptive statistics values oneway one-way analysis of variance anova analysis of variance corr simple correlation among variables twoway scatter produce scatter plot of outcome vs. predictor graph matrix produce multiple twoway scatter plot at a time Scripts in the real Command Window cd E:\MIT\09Spring\STATALAB\DATA (change this part to your own local directory) use hedonic, clear log using log1, text summarize 1) T-test (One sample and two independence samples)2 11.220 Quantitative Reasoning and Statistical Methods for Planning /// Compare the mean of one variable to some constant value μttest size = 1770 * Can we reject H0 = 1770? Why? ttest size = 2000 * Can we reject H0 μ= 2000? Why? ttest size = 1770, level(99) * Can we reject H0 μ= 1770 now? *Note: This is to infer whether the mean of the population equals 1700 or 2000, given the sample mean we already know. /// Compare the mean of two different variables ttest bed = bath, unpaired *Note: Here I use the option “unpaired” since the means are from different variables. “Paired” ttest is by default, which is designed to compare the means of the same variable from different samples. Think of a “pre-post” situation. /// Compare the mean price of old houses vs. new houses tab old, summarize(price) ttest price, by(old) * Can we reject H0:μp_old = μp_new? ttest price, by(old) unequal *Note: If we concern that the samples may have different variances, we need to include “unequal” option. 2) Analysis of Variance /// See whether the old houses and new houses have equal variance? oneway price, old * Check the chi2 value, what do you find? anova price, old 3) Things to do before run into “regression” /// See the simple correlation among variables before we do regression, and this help us to roughly determine which predictors to be included. corr price lot year_b size room bed bath year_s /// Plot the outcome against some predictors graph matrix price lot size room, half /// Plot “price” against “lot” with fitted linear regression line twoway scatter price lot || lfit price lot /// Plot “price” against “lot” with 95% confidence interval twoway scatter price lot || lfitci price lot /// Do simple linear regression! regress price lot Exercises 1: Test whether the μ of lot size = 8600? 8900? On a 95% confidence level. 2: Test whether the μof lot size is statistically different between new and old house. 3. Test whether the variances of price are different for houses with different interior quality? Hint: use “q1” to divide the data into 3 groups. 4. Plot “price” against “room” with fitted regression line and confidence