All events that people believe are the cause of the decline of the outh will be coded as 1, otherwise coded as 0. All things people grew up with will be coded as 1, otherwise 0. Thus, the two variables across each of the societal elements represent a 2x2 contigency table. We predict a negative coefficient for this correlation, meaning the less likely someone experienced something growing up, the more likely they are to think it is a cause of the decline of the youth. This analysis will be run as follows: by thing: phi causedecline hadaskid We will also code the variables for whether they represent a technologly. We will create a histogram of the distribution of phi correlation coefficients. We predict the average coeffecient to be negative, and those that represent technology are more negative than non-technology-based societal elements. After analyzing the data using the Pih coefficients, we learned thatr such coefficients cannot be negative, thus the predicted direction of results cannot be tested. As such, we are re-registering our prediction for directionality. As the dependent variable is what people think is causing a decline in the youth and our predicted 'causing' variable is whether they had it as a child, we will analyze the data using the following code: by thing: probit causedecline hadaskid, vce(robust) 1.14.2020 updtae: As a secondary analysis to get an overall sense of the strength of the data, we will run a form of robust meta-analysis on the individual odds ratios, weighting each by the base-rate of what % people think the specific element is causing a decline in the youth. Thus, elements with very very low base-rates such as 'ballroom dancing ' (1.33% of people think this is causing a decline in the youth) would be weighted very little while 'Social Media' (72.93% think it is causing a decline) would be weighted very heavily. This will be attempted with the following code: robumeta OR, study(study) uweights(baserate) The final analysis plan, that the meta-analysis was not possible, involved combining the above analyses into a within-person analysis using robust standard errors xtmixed causedecline hadaskid i.thing || id:, vce(robust) where we likewise also tested for order effects, and whether conditioning on age changed the results.