How many people clicked product detail page
## [1] "Number of users that checked product detail page: 164"
## [1] "Percentage of users that checked product detail page: 0.987951807228916"
How many purchases people made?
## One_Purchase One_Purchase_Pct Two_Purchases
## More than 60s Users 39 33.05085 78
## More than 30s Users 58 38.66667 91
## Two_Purchases_Pct More_Than_Two More_Than_Two_Pct
## More than 60s Users 66.10169 1 0.8474576
## More than 30s Users 60.66667 1 0.6666667
## Total_Users
## More than 60s Users 118
## More than 30s Users 150
data pattern: How many purchased product’s clinical results are
seen. (Product-Level Analysis)
## === Analysis for >60s Users ===
## Total purchases with claims: 54
## → Clinical image seen: 12 ( 22.22 % )
## → Clinical result seen: 2 ( 3.7 % )
## → Any clinical info seen: 14 ( 25.93 % )
## === Analysis for >30s Users ===
## Total purchases with claims: 64
## → Clinical image seen: 12 ( 18.75 % )
## → Clinical result seen: 2 ( 3.12 % )
## → Any clinical info seen: 14 ( 21.88 % )
Proportion that choose Group I product
##
## === Treatment Analysis for >60s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 98 0.510 0.502 0.0508 0.409
## 2 Treatment 97 0.557 0.499 0.0507 0.456
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = -0.64811, df = 193, p-value = 0.5177
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.1879961 0.0950022
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5102041 0.5567010

##
## === Treatment Analysis for >30s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 122 0.508 0.502 0.0454 0.418
## 2 Treatment 118 0.542 0.500 0.0461 0.451
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = -0.52817, df = 237.78, p-value = 0.5979
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.16164790 0.09329558
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5081967 0.5423729

## Sample size needed per group for >60s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 1805.917
## p1 = 0.5102041
## p2 = 0.556701
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
## Sample size needed per group for >30s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 3350.158
## p1 = 0.5081967
## p2 = 0.5423729
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
Proportion that choose Group I product - first choice
##
## === First-Purchase Treatment Analysis for >60s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 57 0.526 0.504 0.0667 0.393
## 2 Treatment 60 0.55 0.502 0.0648 0.420
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = -0.2547, df = 114.64, p-value = 0.7994
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.2078824 0.1605139
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5263158 0.5500000

##
## === First-Purchase Treatment Analysis for >30s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 73 0.548 0.501 0.0587 0.431
## 2 Treatment 76 0.526 0.503 0.0577 0.411
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = 0.26298, df = 146.79, p-value = 0.7929
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.1409097 0.1841685
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5479452 0.5263158

## Sample size needed per group for >60s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 6954.243
## p1 = 0.5263158
## p2 = 0.55
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
## Sample size needed per group for >30s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 8341.146
## p1 = 0.5479452
## p2 = 0.5263158
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
Treatment effect for those users that have at least seen clinical
image/results once
##
## === Clinical Info Filtered Analysis for >60s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 98 0.510 0.502 0.0508 0.409
## 2 Treatment 32 0.688 0.471 0.0832 0.518
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = -1.8184, df = 55.864, p-value = 0.07437
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.37262691 0.01803508
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5102041 0.6875000

##
## === Clinical Info Filtered Analysis for >30s Users ===
## # A tibble: 2 × 7
## Group n mean_treatment_pa sd_treatment_pa se_treatment_pa ci_lower
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Control 122 0.508 0.502 0.0454 0.418
## 2 Treatment 32 0.688 0.471 0.0832 0.518
## # ℹ 1 more variable: ci_upper <dbl>
##
## Welch Two Sample t-test
##
## data: treatment_pa by Group
## t = -1.8904, df = 51.07, p-value = 0.06438
## alternative hypothesis: true difference in means between group Control and group Treatment is not equal to 0
## 95 percent confidence interval:
## -0.36971077 0.01110422
## sample estimates:
## mean in group Control mean in group Treatment
## 0.5081967 0.6875000

## Sample size needed per group for >60s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 118.7819
## p1 = 0.5102041
## p2 = 0.6875
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
## Sample size needed per group for >30s users:
##
## Two-sample comparison of proportions power calculation
##
## n = 116.207
## p1 = 0.5081967
## p2 = 0.6875
## sig.level = 0.05
## power = 0.8
## alternative = two.sided
##
## NOTE: n is number in *each* group
Loterry choice
## === First Choice Analysis for >60s Users ===
## Proportion of lottery_choice = 1 (Keep the product): 0.1724 ( 17.24 %)
## === First Choice Analysis for >30s Users ===
## Proportion of lottery_choice = 1 (Keep the product): 0.2177 ( 21.77 %)
Check the browsing pattern
##
## === Browsing Behavior Summary for >60s Users ===
## → Checked ingredients text: 18.8 %
## → Checked product reviews: 50.43 %
## → Clicked 'Show More' button: 17.09 %
## → Clicked 'How to Use' button: 5.13 %
## → Viewed clinical image: 13.68 %
## → Viewed clinical result: 2.56 %
##
## === Browsing Behavior Summary for >30s Users ===
## → Checked ingredients text: 14.77 %
## → Checked product reviews: 41.61 %
## → Clicked 'Show More' button: 14.09 %
## → Clicked 'How to Use' button: 4.03 %
## → Viewed clinical image: 10.74 %
## → Viewed clinical result: 2.01 %
Compare the purchased product characteristic
##
## === Group Comparison (First Purchase) for >60s Users ===
## Control.Control Treatment.Treatment Difference.Treatment
## claim_benchmark 4.5714 4.5167 -0.0548
## lottery_binary 0.1607 0.1833 0.0226
## avg_awareness_pa 9.5296 8.9153 -0.6143
## avg_loyalty_pa 5.9030 5.7401 -0.1630
## avg_quality_pa 7.0081 6.9053 -0.1027
## Rating 4.3777 4.4342 0.0566
## Reviews 1014.8772 879.1833 -135.6939
## h_price 54.9123 55.7667 0.8544
## treatment_pa 0.5263 0.5500 0.0237
## p_value
## claim_benchmark 0.8290
## lottery_binary 0.7494
## avg_awareness_pa 0.2227
## avg_loyalty_pa 0.1830
## avg_quality_pa 0.2240
## Rating 0.1225
## Reviews 0.5433
## h_price 0.7809
## treatment_pa 0.7994
##
## === Group Comparison (First Purchase) for >30s Users ===
## Control.Control Treatment.Treatment Difference.Treatment
## claim_benchmark 4.6197 4.5921 -0.0276
## lottery_binary 0.1972 0.2368 0.0397
## avg_awareness_pa 9.7512 9.1301 -0.6211
## avg_loyalty_pa 5.9306 5.7992 -0.1314
## avg_quality_pa 7.0480 6.9483 -0.0997
## Rating 4.3865 4.4364 0.0499
## Reviews 1048.1644 958.6974 -89.4670
## h_price 55.8630 56.5132 0.6501
## treatment_pa 0.5479 0.5263 -0.0216
## p_value
## claim_benchmark 0.9005
## lottery_binary 0.5626
## avg_awareness_pa 0.1831
## avg_loyalty_pa 0.2437
## avg_quality_pa 0.1914
## Rating 0.1191
## Reviews 0.6440
## h_price 0.8141
## treatment_pa 0.7929
##
## === Group Comparison (All Purchases) for >60s Users ===
## Control.Control Treatment.Treatment Difference.Treatment
## claim_benchmark 4.5417 4.5464 0.0047
## lottery_binary 0.1667 0.1753 0.0086
## avg_awareness_pa 8.9447 9.0021 0.0574
## avg_loyalty_pa 5.7456 5.7524 0.0069
## avg_quality_pa 6.9280 6.9348 0.0068
## Rating 4.4427 4.4380 -0.0048
## Reviews 838.8980 735.2680 -103.6299
## h_price 54.7857 57.0103 2.2246
## treatment_pa 0.5102 0.5567 0.0465
## p_value
## claim_benchmark 0.9802
## lottery_binary 0.8749
## avg_awareness_pa 0.8779
## avg_loyalty_pa 0.9408
## avg_quality_pa 0.9133
## Rating 0.8728
## Reviews 0.4749
## h_price 0.3406
## treatment_pa 0.5177
##
## === Group Comparison (All Purchases) for >30s Users ===
## Control.Control Treatment.Treatment Difference.Treatment
## claim_benchmark 4.5966 4.6017 0.0051
## lottery_binary 0.2017 0.2119 0.0102
## avg_awareness_pa 9.1922 9.1206 -0.0715
## avg_loyalty_pa 5.7998 5.7894 -0.0105
## avg_quality_pa 6.9641 6.9579 -0.0063
## Rating 4.4449 4.4443 -0.0007
## Reviews 883.4754 796.0763 -87.3991
## h_price 55.8934 56.9407 1.0472
## treatment_pa 0.5082 0.5424 0.0342
## p_value
## claim_benchmark 0.9762
## lottery_binary 0.8473
## avg_awareness_pa 0.8393
## avg_loyalty_pa 0.9048
## avg_quality_pa 0.9141
## Rating 0.9797
## Reviews 0.5029
## h_price 0.6219
## treatment_pa 0.5979
Among people in the treatment group, do those who purchased products
and saw any of their clinical claims rate the claim’s credibility higher
than those who purchased but didn’t see the claims? - Yes, but not
significantly
##
## === User-Level Claim Benchmark vs. Clinical Seen (Treatment Only) - >60s Users ===
##
## Welch Two Sample t-test
##
## data: claim_benchmark_user by clinical_seen_user
## t = -0.72779, df = 32.852, p-value = 0.4719
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
## -0.7117398 0.3367398
## sample estimates:
## mean in group 0 mean in group 1
## 4.479167 4.666667
##
## Kolmogorov-Smirnov test for distribution difference:
##
## Exact two-sample Kolmogorov-Smirnov test
##
## data: user_level$claim_benchmark_user[user_level$clinical_seen_user == 0] and user_level$claim_benchmark_user[user_level$clinical_seen_user == 1]
## D = 0.1875, p-value = 0.4833
## alternative hypothesis: two-sided
##
## [1] "Group-level summary statistics:"
## clinical_seen_user claim_benchmark_user.mean claim_benchmark_user.sd
## 1 0 4.479167 1.2202125
## 2 1 4.666667 0.6513389
## claim_benchmark_user.n
## 1 48
## 2 12

##
## === User-Level Claim Benchmark vs. Clinical Seen (Treatment Only) - >30s Users ===
##
## Welch Two Sample t-test
##
## data: claim_benchmark_user by clinical_seen_user
## t = -0.36129, df = 29.251, p-value = 0.7205
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
## -0.5895750 0.4124917
## sample estimates:
## mean in group 0 mean in group 1
## 4.578125 4.666667
##
## Kolmogorov-Smirnov test for distribution difference:
##
## Exact two-sample Kolmogorov-Smirnov test
##
## data: user_level$claim_benchmark_user[user_level$clinical_seen_user == 0] and user_level$claim_benchmark_user[user_level$clinical_seen_user == 1]
## D = 0.18229, p-value = 0.5178
## alternative hypothesis: two-sided
##
## [1] "Group-level summary statistics:"
## clinical_seen_user claim_benchmark_user.mean claim_benchmark_user.sd
## 1 0 4.578125 1.2574185
## 2 1 4.666667 0.6513389
## claim_benchmark_user.n
## 1 64
## 2 12

Among all people, do those who purchased products and saw any of
their clinical claims rate the claim’s credibility higher than those who
purchased but didn’t see the claims? - Only slitely higher
##
## === Comparing Treatment Users Who Saw Clinical Info on Purchases vs. All Others - >60s Users ===
##
## Welch Two Sample t-test
##
## data: claim_benchmark_user by group
## t = -0.59197, df = 25.095, p-value = 0.5592
## alternative hypothesis: true difference in means between group Other and group Seen is not equal to 0
## 95 percent confidence interval:
## -0.6172202 0.3415792
## sample estimates:
## mean in group Other mean in group Seen
## 4.528846 4.666667
##
## Kolmogorov-Smirnov test:
##
## Exact two-sample Kolmogorov-Smirnov test
##
## data: compare_df$claim_benchmark_user[compare_df$group == "Other"] and compare_df$claim_benchmark_user[compare_df$group == "Seen"]
## D = 0.21474, p-value = 0.2837
## alternative hypothesis: two-sided
##
## [1] "Summary statistics:"
## # A tibble: 2 × 4
## group mean sd n
## <chr> <dbl> <dbl> <int>
## 1 Other 4.53 1.40 105
## 2 Seen 4.67 0.651 12

##
## === Comparing Treatment Users Who Saw Clinical Info on Purchases vs. All Others - >30s Users ===
##
## Welch Two Sample t-test
##
## data: claim_benchmark_user by group
## t = -0.30057, df = 21.035, p-value = 0.7667
## alternative hypothesis: true difference in means between group Other and group Seen is not equal to 0
## 95 percent confidence interval:
## -0.5278790 0.3945457
## sample estimates:
## mean in group Other mean in group Seen
## 4.600000 4.666667
##
## Kolmogorov-Smirnov test:
##
## Exact two-sample Kolmogorov-Smirnov test
##
## data: compare_df$claim_benchmark_user[compare_df$group == "Other"] and compare_df$claim_benchmark_user[compare_df$group == "Seen"]
## D = 0.21296, p-value = 0.285
## alternative hypothesis: two-sided
##
## [1] "Summary statistics:"
## # A tibble: 2 × 4
## group mean sd n
## <chr> <dbl> <dbl> <int>
## 1 Other 4.6 1.37 137
## 2 Seen 4.67 0.651 12

Missed clinical exposures
##
## === Clinical Exposure Summary for >60s Users ===
## → Number of interactions with claims AND NO clinical info seen: 87
## → Number of interactions with claims AND clinical info seen: 25
## → Proportion seen (seen / (missed + seen)): 22.32 %
##
## === Clinical Exposure Summary for >30s Users ===
## → Number of interactions with claims AND NO clinical info seen: 97
## → Number of interactions with claims AND clinical info seen: 25
## → Proportion seen (seen / (missed + seen)): 20.49 %