I tested possible variations by web site, geographic part, and you can ethnicity playing with t-evaluating and 
you will investigation regarding difference (ANOVA) toward LIWC classification percent. Into a few other sites, half dozen of your own several t-assessment was indeed extreme regarding pursuing the categories: first-people just one [t(3998) = ?5.61, p Secondary Table 2 to own function, fundamental deviations, and contrasts between ethnic organizations). Contrasts found high differences when considering White and all of most other cultural communities inside the five of your half a dozen high ANOVAs. Hence, we incorporated ethnicity given that an effective dummy-coded covariate in analyses (0 = Light, 1 = Any other ethnic groups).
Of one’s several ANOVA screening regarding geographical part, simply a couple were extreme (family relations and you may confident emotion). Since the variations just weren’t commercially important, i did not believe geographic part from inside the further analyses.
Overall performance
Frequency out of term fool around with is evident in detailed analytics (get a hold of Table step one) and you will thru keyword-clouds. The term-affect method depicts the absolute most commonly used conditions across the whole shot and in each of the a long time. The expression-cloud program immediately excludes particular terminology, as well as content (a beneficial, and you may, the) and you will prepositions (so you’re able to, that have, on). The remaining posts terminology try scaled in dimensions relative to its regularity, performing an intuitive portrait quite commonplace stuff conditions all over the latest take to ( Wordle, 2014).
Shape step 1 reveals new 20 typical articles terms and conditions used in the complete decide to try. As well as be seen, probably the most frequently used terms and conditions were like (lookin for the 67% regarding pages), for example (appearing within the 62% away from pages), appearing (appearing into the 55% off profiles), and you can some one (looking during the fifty% from users). Ergo, the preferred terminology was similar across age range.
Shape dos suggests the next 31 most commonly known blogs conditions into the the brand new youngest and earliest a long time. By detatching the original 20 well-known articles words across the shot, we instruct heterogeneity on the matchmaking pages. In the next 29 terms and conditions into youngest age bracket, significant number terminology included get (36% away from users regarding youngest age group), go (33% off users on youngest age bracket), and performs (28% of users on the youngest generation). However, the fresh new oldest age bracket had large percent out-of words including travelling (31% of pages on eldest generation), great (24% of pages about earliest age bracket), and you may relationships (19% regarding profiles about earliest age bracket).
Second 30 common conditions on the youngest and you will earliest many years teams (immediately following deducting brand new 20 typical words from Contour step 1).
Theory Analysis of age Differences in Words in the Relationships Profiles
To check on hypotheses, the brand new percentage of conditions from the matchmaking profile that suit each LIWC category served because based details inside the regressions. I tested many years and you may intercourse as the separate details together with changing to own website and ethnicity.
Hypothesis step 1: Old years would-be of a top portion of terminology on following categories: first-individual plural pronouns, family members, nearest and dearest, health, and positive emotion.
Conclusions mainly served Hypothesis step one (pick Dining table dos). Four of one’s five regressions found a critical main perception having age, such that because the period of this new reputation copywriter increased, the percentage of conditions from the group enhanced throughout the following categories: first-person plural, friends, fitness, and you can self-confident feeling. We receive no significant age impression to your proportion regarding terminology regarding loved ones category.
a sex: 0 (female) and 1 (male). b Web site: The 2 websites was indeed dictomously coded since 1 and you can 0. c Ethnicity: 0 (White) and you may step 1 (Cultural otherwise racial fraction).
a gender: 0 (female) and you can step one (male). b Website: The 2 other sites had been dictomously coded since step one and you will 0. c Ethnicity: 0 (White) and you can 1 (Ethnic otherwise racial fraction).