# load data
dat <- read.csv('/Users/chengnie/Google Drive/Research/UTD/project/SponsorSearch/Jing_Files/Strategic Group - To Jiahui/Data and Analysis/01_digital camera/Cheng/DataH1.txt')
# since firm1 and firm2 are ordered according to the Python code of 36 firm list. There is no problem of permutation of firm pair
dat$FirmPair <- paste(dat$Firm1,dat$Firm2)
# lapply(split(dat,dat$FirmPair),nrow)
x<-sapply(split(dat,dat$FirmPair),nrow)
# x
# data.frame has nicer format in printing
y<-as.data.frame(x[order(x,decreasing=TRUE)])We have 9701 observations for 1275 sessions. We have only 33 unique pairs out of the all potential 36 pairs:
y## x[order(x, decreasing = TRUE)]
## canon bestbuy 1179
## amazon bestbuy 1134
## amazon olympus 948
## canon target 448
## amazon become 430
## amazon target 430
## olympus target 429
## bestbuy become 427
## bestbuy target 423
## canon ebay 337
## amazon ebay 326
## bestbuy ebay 320
## kodak officemax 318
## olympus ebay 313
## kodak samsung 278
## bestbuy samsung 272
## amazon samsung 256
## samsung sony 216
## kodak sony 215
## officemax sony 196
## kodak ecamerafilms 190
## ebay target 142
## nextag sony 133
## nextag samsung 89
## samsung ecamerafilms 73
## become target 53
## nextag philips 43
## kodak philips 39
## philips ecamerafilms 13
## philips officemax 13
## kodak electronicsexpo 10
## ebay samsung 6
## philips sony 2Mean of within group is 0.3633646
Mean of Co-visit is 0.0258736
(table2way <- xtabs(~Co.visit + within.group, data = dat))## within.group
## Co.visit 0 1
## 0 6101 3349
## 1 75 176# http://www.statmethods.net/stats/frequencies.html
summary(table2way) # chi-square test of indepedence## Call: xtabs(formula = ~Co.visit + within.group, data = dat)
## Number of cases in table: 9701
## Number of factors: 2
## Test for independence of all factors:
## Chisq = 127.12, df = 1, p-value = 1.746e-29chisq.test(table2way)##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: table2way
## X-squared = 125.6279, df = 1, p-value < 2.2e-16sapply(split(dat$Co.visit,dat$within.group),mean)## 0 1
## 0.01214378 0.04992908The result shows that the co-visit ratio is higher for within-group pair.
\(CoVisit = \beta_0 + \beta_1 * Within + \epsilon\)
# http://www.ats.ucla.edu/stat/r/dae/logit.htm
logit1 <- glm(Co.visit ~ within.group, data = dat, family = "binomial")
summary(logit1)##
## Call:
## glm(formula = Co.visit ~ within.group, family = "binomial", data = dat)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.3201 -0.3201 -0.1563 -0.1563 2.9702
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.3987 0.1162 -37.86 <2e-16 ***
## within.group 1.4528 0.1396 10.41 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2330.0 on 9700 degrees of freedom
## Residual deviance: 2208.8 on 9699 degrees of freedom
## AIC: 2212.8
##
## Number of Fisher Scoring iterations: 7\(CoVisit = \beta_0 + \beta_1 * Within + \alpha_{FirmPair} + \epsilon\)
logit2 <- glm(Co.visit ~ within.group + FirmPair, data = dat, family = "binomial")
summary(logit2)##
## Call:
## glm(formula = Co.visit ~ within.group + FirmPair, family = "binomial",
## data = dat)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.4838 -0.2880 -0.1845 -0.1189 3.4818
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.6681 0.5023 -9.293 < 2e-16 ***
## within.group 0.6979 0.7121 0.980 0.327079
## FirmPairamazon bestbuy 1.2790 0.5192 2.464 0.013755 *
## FirmPairamazon ebay 1.8839 0.5347 3.523 0.000427 ***
## FirmPairamazon olympus 0.6034 0.5621 1.074 0.283030
## FirmPairamazon samsung 0.5250 0.7116 0.738 0.460617
## FirmPairamazon target 1.0011 0.5521 1.813 0.069777 .
## FirmPairbecome target -14.8979 1477.1774 -0.010 0.991953
## FirmPairbestbuy become 0.2326 0.6743 0.345 0.730167
## FirmPairbestbuy ebay 0.8084 0.5787 1.397 0.162442
## FirmPairbestbuy samsung 0.1720 0.7677 0.224 0.822680
## FirmPairbestbuy target 1.2445 0.5438 2.289 0.022107 *
## FirmPaircanon bestbuy 0.9147 0.5388 1.698 0.089533 .
## FirmPaircanon ebay 0.8150 0.6311 1.291 0.196558
## FirmPaircanon target -0.7390 0.8687 -0.851 0.394905
## FirmPairebay samsung -14.8979 4390.3074 -0.003 0.997292
## FirmPairebay target -0.9785 1.1233 -0.871 0.383720
## FirmPairkodak ecamerafilms -14.8979 780.1783 -0.019 0.984765
## FirmPairkodak electronicsexpo -14.8979 3400.7175 -0.004 0.996505
## FirmPairkodak officemax -14.8979 603.0553 -0.025 0.980291
## FirmPairkodak philips -15.5958 1722.0203 -0.009 0.992774
## FirmPairkodak samsung -0.9570 0.8708 -1.099 0.271807
## FirmPairkodak sony -1.3957 1.1222 -1.244 0.213619
## FirmPairnextag philips -14.8979 1639.9717 -0.009 0.992752
## FirmPairnextag samsung 0.1908 1.1241 0.170 0.865218
## FirmPairnextag sony -0.2147 1.1225 -0.191 0.848340
## FirmPairofficemax sony -14.8979 768.1439 -0.019 0.984526
## FirmPairolympus ebay 0.3211 0.7110 0.452 0.651567
## FirmPairolympus target -1.3910 1.1201 -1.242 0.214309
## FirmPairphilips ecamerafilms -14.8979 2982.6266 -0.005 0.996015
## FirmPairphilips officemax -14.8979 2982.6266 -0.005 0.996015
## FirmPairphilips sony -15.5958 7604.2355 -0.002 0.998364
## FirmPairsamsung ecamerafilms -14.8979 1258.6621 -0.012 0.990556
## FirmPairsamsung sony NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2330 on 9700 degrees of freedom
## Residual deviance: 2090 on 9668 degrees of freedom
## AIC: 2156
##
## Number of Fisher Scoring iterations: 18However, \(\beta_1\) becomes insignificant after we control for FirmPair fixed effect.