We checked out possible distinctions because of the webpages, geographical part, and ethnicity playing with t-testing and you will investigation out-of difference (ANOVA) to your LIWC category percent. On the two websites, half a dozen of a dozen t-screening was basically extreme regarding the pursuing the groups: first-person only 1 [t(3998) = ?5.61, p Supplementary Dining table dos getting means, simple deviations, and you will contrasts between cultural teams). Contrasts shown significant differences between White as well as almost every other ethnic organizations in the four of your six significant ANOVAs. For this reason, i provided ethnicity since the an effective dummy-coded covariate inside the analyses (0 = White, 1 = Another cultural organizations).
Of your a dozen ANOVA evaluation connected with geographic part, just several was high (family members and you may self-confident emotion). Given that distinctions just weren’t technically meaningful, we failed to envision geographical region in subsequent analyses.
Performance
Regularity from term use is evident for the detailed analytics (pick Table 1) and you can through term-clouds. The word-cloud technique portrays the essential widely used terms along the whole test along with each of the age range. The expression-cloud program instantly excludes certain terms, together with content (an effective, and you may, the) and you will prepositions (to, having, on). The rest stuff terms are scaled in size in accordance with the volume, performing an user-friendly portrait of the very most common stuff conditions all over the latest decide to try ( Wordle, 2014).
Profile step 1 shows the fresh 20 most common stuff terms utilized in the complete take to. As well as get noticed, many commonly used terms and conditions have been like (looking inside the 67% out-of users), for example (searching during the 62% away from users), appearing (looking when you look at the 55% out-of users), and you may some body (lookin inside the fifty% off users). For this reason, widely known terms was comparable across the a long time.
Figure 2 reveals another 30 most typical content terminology in the new youngest and you can eldest age range. By eliminating the initial 20 common stuff conditions along side attempt, i show heterogeneity regarding relationship pages. Next 30 words to your youngest generation, significant number terms included rating (36% out-of pages throughout the youngest age group), go (33% regarding profiles about youngest age bracket), and works (28% out of users on youngest age group). Having said that, the newest eldest age bracket got higher percentages out of conditions instance traveling (31% out-of profiles from the eldest generation), high (24% regarding users on the earliest age group), and you will relationships (19% out of profiles regarding oldest generation).
Second 31 most frequent conditions throughout the youngest and you may earliest decades teams (immediately after deducting this new 20 most commonly known words regarding Profile 1).
Hypothesis Testing of age Differences in Language during the Matchmaking Users
To check hypotheses, the newest percentage of words in the matchmaking profile that fit for each and every LIWC class supported as the established parameters into the regressions. I examined many years and you may gender because independent variables and modifying for site and ethnicity.
Theory 1: Elderly years might possibly be associated with the a high percentage of words on following the classes: first-person plural pronouns, family, friends, wellness, and self-confident feeling.
Results mainly supported Theory 1 (get a hold of Dining table 2). Five of the five regressions shown a life threatening fundamental perception getting many years, such that because age this new character creator increased, the portion of terminology from the class enhanced on the following categories: first-people plural, nearest and dearest, fitness, and you can confident feelings. We discover no extreme age feeling on the ratio from terminology from the family class.
good Gender: 0 (female) and you will step 1 (male). b Website: The two other sites was dictomously coded once the step one and 0. c Ethnicity: 0 (White) and you may step 1 (Ethnic otherwise racial fraction).
good Gender: 0 (female) and you can step 1 (male). b Webpages: The 2 other sites was indeed dictomously coded as the step 1 and 0. c Ethnicity: 0 (White) and you will step 1 (Ethnic or racial fraction).