Within for each and every website, i amassed 1,one hundred thousand profiles of heterosexual boys and you can step one,100 profiles out of heterosexual people. Users search for pages via geographic venue, ages, and you can intercourse strain.
To make certain a geographical dispersion out of pages, we chosen equal quantities of pages regarding four big metropolises also metropolitan, suburban, and you will rural section: La, Denver, il, Atlanta, and you will Nyc. We at random chose zero rules off each one of the four components to find users. In this per zip code, per intercourse, i then at random selected profiles certainly five age range: early younger adulthood (old 18–29 decades; letter = 1,000; M = , SD = step three.17), later younger adulthood/early midlife (aged 31–44 many years; n = 1,000; M = , SD = 5.61), later midlife (old 50–64 decades; n = 1,000; Meters = , SD = 3.99), and you may the elderly (aged more than 65 years; n = step one,000; Yards = , SD = cuatro.29). I utilized these stratifications to assure the full a long time regarding relationships profiles inside the testing. Due to the fact the elderly category you’ll utilize around three decades, we handled age while the a continuing variable in place of while the a beneficial group changeable in the analyses.
From per reputation, we extracted: intercourse, ages, ethnicity, while the “In the Myself” otherwise “Inside my Words” totally free response area. To assure anonymity out-of reputation writers, we don’t get additional group recommendations (e.g., training, religious tastes, income) that could act as identifying guidance. The fresh new sampling system is depicted for the Additional Appendix A good .
Brand new sample ranged when you look at the age regarding 18 to 95 many years. A different t-sample shown no difference in imply age for women (Meters = , SD = ) and you can people (Meters = , SD = ). The post on ethnicity regarding the test is actually 70% White/Caucasian, 11% Black/African american, 7% Hispanic/Latino, 2% Far eastern, and 10% mixed battle/other.
Profile stuff
We utilized the LIWC app to analyze the content of one’s profiles. This software calculates brand new regularity and you will proportions of certain types of terms and conditions contained in this a text document. This new LIWC system compares for each word-of a book file with an internal dictionary in excess of cuatro,five-hundred words allotted to phrase classes. This study drew to your eleven established LIWC categories: first-people one pronouns, first-individual plural pronouns, family relations, family relations, performs, conclusion, currency, health, sexuality, positive feelings, and you can negative feeling. Dining table step one contains 
We and additionally composed a group of conditions to possess appeal not available for the oriented LIWC classes. We implemented measures to possess construction out of LIWC classes ( Tausczik & Pennebaker, 2010) by generating a thorough set of conditions out-of dictionaries, thesauruses, forms out-of past search, and you can words produced by the study party. Then, we picked 25 terms and conditions really affiliate away from appeal according to appearance in thesauruses and new member solutions (elizabeth.grams., cute, good-searching, handsome, hot). New elegance class is actually almost completely distinctive from brand new sexual group, with just one overlapping phrase (sexy). Examples on attractiveness category are used in Desk step 1; to your over list of terms and conditions from the attractiveness classification, get a hold of Second Desk step 1 .
Analytic Approach
We very first tested detailed statistics to your proportions of terminology someone of various age included in its relationship users. We and made illustrative figures portraying the most popular conditions.
I following considered hypothesis assessment using average minimum squares regression. The outcomes parameters in this study was basically the fresh new proportion away from terms and conditions suitable each one of the 12 groups in the LIWC analyses. The new LIWC kinds had been most of the undoubtedly skewed because of the number of no opinions (i.e., new member did not have fun with any terms and conditions throughout the classification). I ran for every single data having a rectangular-supply conversion (regularly address non-normality inside the early in the day education using the LIWC; DeAndrea, Shaw & Levine, 2010; Hirsh & Peterson, 2009). The newest development away from conclusions try comparable after using the changes. For easy interpretation, findings is actually displayed by using the untransformed LIWC group studies. New separate variable is actually age, treated just like the a continuing adjustable. We along with included gender.