Skip to main content
. 2017 Sep 1;5(3):197–212. doi: 10.1089/big.2017.0074

Table 2.

The amount of Likes one needs to remove (effort) to cloak different users' traits predicted by the LRSVD models

  Inline graphic Inline graphic
Task All TP FP All, % TP, % FP, %
Age Inline graphic37 10.3 13.0 5.8 7.7 9.7 4.4
Agreeableness Inline graphic5 5.0 6.5 5.0 2.3 3.3 2.3
Conscientiousness Inline graphic5 4.7 6.7 4.7 3.9 4.7 3.9
Extraversion Inline graphic5 4.4 5.9 4.3 1.9 2.4 1.8
IQ Inline graphic90 6.9 16.3 4.6 4.5 9.0 3.5
IQ Inline graphic130 6.6 3.4 7.3 2.8 3.5 2.6
Is democrat 8.5 8.5 2.0 1.7 1.7 0.3
Is drinking 6.8 7.5 3.9 2.0 2.2 1.2
Is female 10.0 10.0 5.5 1.9 1.9 1.3
Is gay 5.7 10.9 3.2 3.8 7.4 2.2
Is homosexual 3.5 6.6 2.9 2.4 4.7 1.9
Is lesbian 3.1 5.4 2.8 1.9 3.5 1.7
Is Muslim 11.7 27.8 2.9 9.6 20.2 3.9
Is single 13.7 15.5 7.9 3.4 3.8 2.1
Is smoking 8.4 9.8 5.6 2.8 3.2 1.9
Life satisfaction Inline graphic6 5.1 7.2 4.6 2.2 3.2 2.0
Network density Inline graphic65 10.5 15.3 10.4 2.1 2.6 2.1
Neuroticism Inline graphic5 9.1 5.7 9.2 2.2 1.6 2.2
Num friends Inline graphic585 5.0 6.6 4.2 2.1 2.5 1.9
Openness Inline graphic5 6.7 7.7 6.6 2.3 2.8 2.3
ss belief = 1 5.7 6.9 4.9 2.9 3.6 2.4
ss belief = 5 8.3 11.1 7.8 2.1 2.5 2.1
Uses drugs 12.2 12.1 12.2 2.7 3.3 2.2
Mean 7.5 9.9 5.6 3.1 4.5 2.3
Median 6.8 7.7 4.9 2.3 3.3 2.1

Absolute efforts (numbers of Likes removed) are presented in the left panel, and relative efforts (percentages of Likes removed) are in the right panel. For each panel, we show in the first column, the full set of users with data for the trait, in the second column only the TP users, and in the third column only the FP users.

FP, false positive; LRSVD, logistic regression singular-value decomposition; TP, true positive.