I tested prospective differences by website, geographical part, and ethnicity playing with t-assessment and analysis from difference (ANOVA) towards LIWC group rates. On several other sites, six of your several t-screening was in fact extreme throughout the following the classes: first-people just one [t(3998) = ?5.61, p Secondary Desk dos to possess means, important deviations, and you will contrasts between ethnic groups). Contrasts shown significant differences when considering Light and all sorts of most other cultural communities from inside the five of one’s half dozen high ANOVAs. Thus, we integrated ethnicity given that a beneficial dummy-coded covariate into the analyses (0 = White, step one = Any ethnic groups).
Of the a dozen ANOVA testing connected with geographical region, just two was high (friends and you may confident feeling). Since variations weren’t officially important, we didn’t consider geographical area from inside the further analyses.
Performance
Regularity out-of phrase explore is evident in the detailed analytics (come across Table step 1) and you can thru word-clouds. The term-cloud approach portrays the quintessential widely used words over the whole try and in each one of the a long time. The term-cloud system automatically excludes certain terminology, as well as blogs (an excellent, and you can, the) and you may prepositions (so you’re able to, with, on). The remaining articles words is actually scaled in size in accordance with its regularity, starting an intuitive portrait really prevalent stuff terms all over brand new decide to try ( Wordle, 2014).
Profile step one shows the fresh 20 most frequent articles terms and conditions included in the entire sample. As well as be seen, the quintessential frequently used words had been love (looking when you look at the 67% away from pages), eg (looking within the 62% from profiles), lookin (appearing for the 55% regarding users), and you can some one (appearing from inside the 50% away from profiles). Therefore, widely known words was equivalent round the age groups.
Shape dos suggests another 31 typical stuff terms and conditions in the youngest and you will earliest age range. By removing the original 20 well-known posts conditions over the take to, i train heterogeneity in the relationship users. Next 30 terms and conditions to the youngest generation, raised percentage terms incorporated get (36% out of profiles on youngest age bracket), go (33% out of users from the youngest age bracket), and you may functions (28% from profiles in the youngest age bracket). Alternatively, the newest eldest age bracket got high rates regarding terminology such travel (31% off profiles throughout the eldest age group), high (24% from users on the eldest generation), and you can relationships (19% away from profiles on the oldest age group).
Next 30 typical terms and conditions regarding youngest and you may eldest ages communities (shortly after subtracting the fresh 20 common terms and conditions out-of Contour step one).
Theory Testing old Differences in Language in the Relationships Profiles
To check on hypotheses, this new portion of words throughout the relationship profile that fit for every LIWC group served because oriented variables inside regressions. I looked at many years and you will gender once the separate variables and additionally adjusting having webpages and ethnicity.
Theory step one: Old ages would be of the increased portion of terms and conditions regarding after the categories: first-individual plural pronouns, household members, nearest and dearest, health, and you can self-confident feeling.
Conclusions largely served Hypothesis step one (select Dining table 2). Five press this link of your four regressions found a significant fundamental effect for age, in a way that as the ages of the fresh new reputation blogger increased, the fresh new part of terms about category enhanced in the pursuing the categories: first-individual plural, loved ones, health, and you may positive emotion. We discovered no significant many years perception to the proportion out-of terms throughout the relatives category.
a great Gender: 0 (female) and you can 1 (male). b Webpages: The 2 websites was in fact dictomously coded as the step 1 and 0. c Ethnicity: 0 (White) and step one (Cultural or racial minority).
a Gender: 0 (female) and you can 1 (male). b Webpages: The 2 websites was dictomously coded because step one and you can 0. c Ethnicity: 0 (White) and you may step one (Ethnic otherwise racial minority).