We checked out potential distinctions of the website, geographical part, and you can ethnicity having fun with t-screening and you can research off variance (ANOVA) into the LIWC category percent. Toward two websites, half dozen of the twelve t-assessment had been extreme about following the groups: first-people singular [t(3998) = ?5.61, p Supplementary Desk dos for mode, standard deviations, and you may contrasts anywhere between cultural communities). Contrasts revealed significant differences when considering White and all sorts of most other ethnic organizations for the five of your half a dozen high ANOVAs. Ergo, we provided ethnicity since the an effective dummy-coded covariate from inside the analyses (0 = Light, step 1 = Other ethnic organizations).
Of your own 12 ANOVA tests about geographic part, simply a couple of was extreme (members of the family and you can self-confident feeling). As variations just weren’t theoretically important, we failed to imagine geographical region in then analyses.
Efficiency
Volume of word fool around with is evident from inside the descriptive statistics (find Dining table step one) and via term-clouds. The phrase-affect technique illustrates many popular terms and conditions along the whole try along with all the a long time. The expression-cloud program automatically excludes specific terms and conditions, and blogs (a good, and you will, the) and you will prepositions (to help you, that have, on). The rest stuff terminology was scaled in dimensions according to the frequency, starting an user-friendly portrait of the very most prevalent posts terms and conditions round the the attempt ( Wordle, 2014).
Shape 1 reveals the newest 20 most frequent articles terms utilized in the entire take to. As well as rise above the crowd, one particular commonly used terms and conditions had been love (lookin when you look at the 67% of profiles), such as for instance (lookin within the 62% of pages), searching (appearing for the 55% from profiles), and you may some body (lookin inside the fifty% off profiles). Thus, the most common conditions have been comparable across a long time.
Profile dos shows another 31 most commonly known posts conditions inside the new youngest and you can oldest a long time. By detatching the initial 20 preferred content terms and conditions along side decide to try, i illustrate heterogeneity throughout the dating users. Within the chat avenue-sovellus next 30 words to your youngest generation, high level percentage conditions included score (36% from users on the youngest age bracket), go (33% away from profiles in the youngest age group), and works (28% of pages regarding the youngest generation). Conversely, brand new earliest age bracket got large percent of words such travel (31% off users in the earliest age bracket), great (24% out-of users regarding the oldest generation), and you may matchmaking (19% off pages on the oldest age bracket).
Second 31 common terms and conditions regarding the youngest and you will eldest many years groups (immediately after subtracting new 20 most commonly known terminology regarding Figure step one).
Theory Analysis of age Variations in Vocabulary for the Relationship Pages
To evaluate hypotheses, the brand new part of terminology on relationship character that fit for each and every LIWC category served as oriented parameters into the regressions. We checked-out decades and you may gender given that separate parameters also changing for webpages and you can ethnicity.
Theory step 1: Elderly ages was for the increased part of words regarding pursuing the classes: first-people plural pronouns, family members, friends, wellness, and you may positive feelings.
Conclusions mostly served Theory step one (look for Dining table 2). Five of your own five regressions found a significant head impression to own years, such that because ages of the latest profile creator increased, brand new portion of terms and conditions throughout the classification improved regarding after the categories: first-person plural, loved ones, health, and self-confident feeling. We located no significant decades perception for the proportion away from terminology throughout the relatives class.
good Gender: 0 (female) and you can 1 (male). b Web site: The 2 websites was indeed dictomously coded due to the fact step one and 0. c Ethnicity: 0 (White) and you can step 1 (Ethnic or racial fraction).
good Gender: 0 (female) and step one (male). b Web site: The 2 websites had been dictomously coded because step one and you can 0. c Ethnicity: 0 (White) and you will step one (Ethnic or racial fraction).