- see [[240201_120647 more models|ed user partition v1 (first version)]] # network clustering ![[20240229104250.png]] ```python Partition 0: red # retweeters Partition 1: green # retweeters Partition 2: blue # original sharers everyone else retweets Partition 3: yellow Partition 4: magenta # original sharers? but a bit mixed?? Partition 5: cyan # original sharers Partition 6: orange # original sharers Partition 7-15: white # ignore ``` inter-partition edge count matrix ![[20240229162338 2.png]] ```r # no. of users in each partition part N <char> <int> 1: 0 141120 2: 1 80044 3: 2 17143 4: 3 10750 5: 4 48413 6: 5 10201 7: 6 3974 8: 7 202792 # 7 to 15 (and all users who don't belong to ed's network) ``` # different user subpopulations - all users: 514,437 (users not in Ed's network are assigned to the 'null' partition) - users in Ed's network/partitions: 399,126 - users who shared anything at all during campaign: 233,267 - users in Ed's network & shared anything at all during campaign: 224,718 - users who shared anything at all pre-campaign: 266,551 - users in Ed's network & shared anything at all pre-campaign: 254,911 - users who shared anything at all pre-campaign & during campaign: 230,720 - users who shared anything at all pre-campaign & during campaign & in Ed's network: 223,377 # model specification - condition (centered, -0.5, 0.5) - t0 (centered): pre-treatment sharing - user partition (2 levels) - partitions 0 and 1 (coded 0) - other partitions (coded 1) - narrative cluster (2 levels) - cluster 1 (coded 0) - other clusters (coded 1) Ignore the suffix `C` in the predictor names! ```r # fix effect block and day feols(t1bin ~ conditionC * (t0binSC + partitionC * clusterC) | block + day, d3, cluster = "block") ``` # results See first row (`term == 'conditionC'`) for treatment effect (negative = less likely to share bad content), specifically for user partitions 0 and 1 and narrative cluster 1. - p-vals are < .05 for all models and range from .01 to .04. - `control_dv_raw_1`: mean(raw DV) for control group - `control_dv_pred_2`: mean(DV based on model predictions) for control group (note there's no intercept term for fixed-effects models) - `perc_change_1`: estimate expressed as perc change (denominator is `control_dv_raw_1`) - `perc_change_2`: estimate expressed as perc change (denominator is `control_dv_pred_2`) ## all users: 514,437 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00045 0.00021 -2.10 0.0361 7202118 0.02 0.03 -2.17 -1.55 2: t0binSC 0.15667 0.00089 176.28 0.0000 7202118 0.02 0.03 763.34 543.06 3: partitionC 0.00233 0.00013 18.04 0.0000 7202118 0.02 0.03 11.34 8.07 4: clusterC 0.02807 0.00036 76.93 0.0000 7202118 0.02 0.03 136.78 97.31 5: partitionC:clusterC -0.02125 0.00050 -42.09 0.0000 7202118 0.02 0.03 -103.52 -73.65 6: conditionC:t0binSC -0.00118 0.00113 -1.04 0.2965 7202118 0.02 0.03 -5.73 -4.08 7: conditionC:partitionC 0.00030 0.00022 1.37 0.1721 7202118 0.02 0.03 1.44 1.02 8: conditionC:clusterC 0.00057 0.00056 1.00 0.3168 7202118 0.02 0.03 2.76 1.96 9: conditionC:partitionC:clusterC -0.00034 0.00084 -0.40 0.6890 7202118 0.02 0.03 -1.64 -1.17 ``` ## users in Ed's network/partitions: 399,126 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00052 0.00025 -2.06 0.0397 5587764 0.03 0.04 -1.91 -1.41 2: t0binSC 0.15689 0.00087 180.00 0.0000 5587764 0.03 0.04 573.99 424.66 3: partitionC 0.00031 0.00018 1.75 0.0800 5587764 0.03 0.04 1.13 0.83 4: clusterC 0.02801 0.00036 78.59 0.0000 5587764 0.03 0.04 102.48 75.82 5: partitionC:clusterC -0.01600 0.00065 -24.77 0.0000 5587764 0.03 0.04 -58.53 -43.30 6: conditionC:t0binSC -0.00124 0.00115 -1.08 0.2796 5587764 0.03 0.04 -4.55 -3.37 7: conditionC:partitionC 0.00034 0.00029 1.18 0.2388 5587764 0.03 0.04 1.23 0.91 8: conditionC:clusterC 0.00058 0.00054 1.08 0.2809 5587764 0.03 0.04 2.14 1.58 9: conditionC:partitionC:clusterC -0.00035 0.00106 -0.33 0.7428 5587764 0.03 0.04 -1.27 -0.94 ``` ## users who shared anything at all during campaign: 233,267 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00108 0.00045 -2.41 0.0158 3265738 0.04 0.06 -2.56 -1.72 2: t0binSC 0.15386 0.00086 179.26 0.0000 3265738 0.04 0.06 366.05 245.31 3: partitionC 0.00198 0.00030 6.71 0.0000 3265738 0.04 0.06 4.72 3.16 4: clusterC 0.05900 0.00069 85.94 0.0000 3265738 0.04 0.06 140.36 94.06 5: partitionC:clusterC -0.02976 0.00105 -28.27 0.0000 3265738 0.04 0.06 -70.80 -47.45 6: conditionC:t0binSC -0.00157 0.00124 -1.27 0.2048 3265738 0.04 0.06 -3.74 -2.51 7: conditionC:partitionC 0.00089 0.00049 1.82 0.0691 3265738 0.04 0.06 2.12 1.42 8: conditionC:clusterC 0.00132 0.00104 1.27 0.2056 3265738 0.04 0.06 3.14 2.11 9: conditionC:partitionC:clusterC -0.00095 0.00174 -0.55 0.5851 3265738 0.04 0.06 -2.26 -1.52 ``` ## users in Ed's network & shared anything at all during campaign: 224,718 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00114 0.00046 -2.49 0.0129 3146052 0.04 0.06 -2.62 -1.76 2: t0binSC 0.15387 0.00085 180.02 0.0000 3146052 0.04 0.06 352.68 237.20 3: partitionC 0.00135 0.00031 4.30 0.0000 3146052 0.04 0.06 3.08 2.07 4: clusterC 0.05899 0.00068 86.21 0.0000 3146052 0.04 0.06 135.21 90.94 5: partitionC:clusterC -0.02704 0.00110 -24.58 0.0000 3146052 0.04 0.06 -61.98 -41.68 6: conditionC:t0binSC -0.00173 0.00125 -1.38 0.1675 3146052 0.04 0.06 -3.96 -2.67 7: conditionC:partitionC 0.00095 0.00051 1.85 0.0642 3146052 0.04 0.06 2.18 1.46 8: conditionC:clusterC 0.00139 0.00103 1.35 0.1773 3146052 0.04 0.06 3.20 2.15 9: conditionC:partitionC:clusterC -0.00108 0.00181 -0.60 0.5498 3146052 0.04 0.06 -2.48 -1.67 ``` ## users who shared anything at all pre-campaign: 266,551 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00095 0.00039 -2.44 0.0148 3731714 0.04 0.05 -2.51 -1.74 2: t0binSC 0.14699 0.00083 176.53 0.0000 3731714 0.04 0.05 387.63 268.14 3: partitionC 0.00201 0.00025 7.95 0.0000 3731714 0.04 0.05 5.31 3.67 4: clusterC 0.05108 0.00061 84.24 0.0000 3731714 0.04 0.05 134.71 93.19 5: partitionC:clusterC -0.02977 0.00093 -31.99 0.0000 3731714 0.04 0.05 -78.51 -54.31 6: conditionC:t0binSC -0.00163 0.00117 -1.39 0.1649 3731714 0.04 0.05 -4.29 -2.97 7: conditionC:partitionC 0.00070 0.00042 1.66 0.0970 3731714 0.04 0.05 1.84 1.28 8: conditionC:clusterC 0.00098 0.00093 1.05 0.2920 3731714 0.04 0.05 2.58 1.78 9: conditionC:partitionC:clusterC -0.00054 0.00155 -0.35 0.7273 3731714 0.04 0.05 -1.42 -0.99 ``` ## users in Ed's network & shared anything at all pre-campaign: 254,911 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00101 0.00041 -2.49 0.0128 3568754 0.04 0.06 -2.54 -1.76 2: t0binSC 0.14718 0.00083 177.49 0.0000 3568754 0.04 0.06 371.08 257.47 3: partitionC 0.00138 0.00027 5.10 0.0000 3568754 0.04 0.06 3.48 2.41 4: clusterC 0.05099 0.00060 84.53 0.0000 3568754 0.04 0.06 128.57 89.20 5: partitionC:clusterC -0.02699 0.00098 -27.55 0.0000 3568754 0.04 0.06 -68.05 -47.21 6: conditionC:t0binSC -0.00174 0.00119 -1.47 0.1429 3568754 0.04 0.06 -4.38 -3.04 7: conditionC:partitionC 0.00074 0.00045 1.67 0.0954 3568754 0.04 0.06 1.87 1.30 8: conditionC:clusterC 0.00103 0.00092 1.12 0.2631 3568754 0.04 0.06 2.59 1.79 9: conditionC:partitionC:clusterC -0.00067 0.00162 -0.41 0.6809 3568754 0.04 0.06 -1.68 -1.17 ``` ## users who shared anything at all pre-campaign & during campaign: 230,720 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00106 0.00045 -2.35 0.0188 3230080 0.04 0.06 -2.50 -1.67 2: t0binSC 0.15351 0.00086 179.32 0.0000 3230080 0.04 0.06 361.48 242.33 3: partitionC 0.00190 0.00030 6.33 0.0000 3230080 0.04 0.06 4.47 3.00 4: clusterC 0.05949 0.00069 85.95 0.0000 3230080 0.04 0.06 140.08 93.91 5: partitionC:clusterC -0.02986 0.00106 -28.04 0.0000 3230080 0.04 0.06 -70.31 -47.13 6: conditionC:t0binSC -0.00156 0.00124 -1.25 0.2096 3230080 0.04 0.06 -3.67 -2.46 7: conditionC:partitionC 0.00087 0.00050 1.74 0.0816 3230080 0.04 0.06 2.04 1.37 8: conditionC:clusterC 0.00128 0.00105 1.22 0.2214 3230080 0.04 0.06 3.02 2.03 9: conditionC:partitionC:clusterC -0.00091 0.00176 -0.52 0.6053 3230080 0.04 0.06 -2.14 -1.44 ``` ## users who shared anything at all pre-campaign & during campaign & in Ed's network: 223,377 ```r term estimate std.error statistic p.value nobs control_dv_raw_1 control_dv_pred_2 perc_change_1 perc_change_2 <char> <num> <num> <num> <num> <num> <num> <num> <num> <num> 1: conditionC -0.00112 0.00046 -2.43 0.0149 3127278 0.04 0.07 -2.57 -1.72 2: t0binSC 0.15356 0.00085 179.84 0.0000 3127278 0.04 0.07 350.38 235.38 3: partitionC 0.00136 0.00031 4.31 0.0000 3127278 0.04 0.07 3.10 2.08 4: clusterC 0.05946 0.00069 86.12 0.0000 3127278 0.04 0.07 135.67 91.14 5: partitionC:clusterC -0.02726 0.00111 -24.63 0.0000 3127278 0.04 0.07 -62.19 -41.78 6: conditionC:t0binSC -0.00171 0.00125 -1.36 0.1729 3127278 0.04 0.07 -3.90 -2.62 7: conditionC:partitionC 0.00093 0.00052 1.80 0.0714 3127278 0.04 0.07 2.12 1.43 8: conditionC:clusterC 0.00135 0.00104 1.30 0.1932 3127278 0.04 0.07 3.09 2.08 9: conditionC:partitionC:clusterC -0.00106 0.00182 -0.58 0.5600 3127278 0.04 0.07 -2.42 -1.63 ```