Search Results
: N of percentages for linear mixed-effects model: 350; logLik = −1032; deviance = 2058, REML deviance = 2064. Fixed effects: mean for 1990 = 7.4% ( SE = 0.8, t = 8.9); slope for 1975–1990 = 0.35 ( SE = 0.09, t = 3.9); slope for 1990
if authors from nationalistic and non-nationalistic countries are randomly distributed, the positive and negative errors in the estimated regressions will cancel each other out. In addition, we will include country and year fixed effects to control
Zone 1 Zone 2 Zone 3 Fixed effects Random effects
consider the first enrollment demand with fixed-effects model, given by (1) and taking logarithms (2) where y it is the demand of region i in period t ; x it is a vector of dimension K of explanatory variables ( x it reports the observation of
. Since it is well-accepted that the propensity to patent inventions differs across industries (Cohen et al. 2000 ; Levin et al. 1987 ), additional negative binomial regressions are run but this time with category fixed effects. Table 3 supports the
-of-sample prediction. We begin by regressing IC rates on the log-number of publications, allowing for fixed effects for each year and country in our sample. Denoting countries by c and years by t , we estimate: 3 The first term on the right hand side captures time
