It is interesting to note that the orthogonal pulse in Figure 2 has only β coefficients that differ from a constant.
Only correlation coefficients that matched a p-value <0.02 are displayed in the maps.
However, running the models with 'ReLogit' changed only slightly the coefficients that were obtained (only at the 3rd decimal).Thus, I decided to not use ReLogit given the rather small differences in the outcomes and the constraints imposed by this procedure on calculating predicted probabilities.
One disadvantage of the fixed effects model is that only coefficients for variables that vary over time can be estimated.
Only the correlation coefficients that passed through a t-test with a significance level of 95% are plotted on the world map.
The mean model has a better explanatory power than the OLS model (not shown), but has only seven significant coefficients that produce just two inconsistencies.
Columns (2), (5), (8) include all ( {f}_i^{prime } ), f i ′ × t and f i ′ × t 2, but only coefficients on those that are also "double selected" are shown (a full regression table is available from the authors).
Columns (1), (3), (5) include all f i ′, but only coefficients on those that are also "double selected" are shown (a full regression table is available from the authors) *p < .10 **p < .05 ***p < .01.01
Columns (1), (3), (5) include all f i ′, but only coefficients on those that are also "double selected" are shown (full regressions table are available from the authors) *p < .10 **p < .05 ***p < .01.01
Although, we are ultimately only interested in the CA coefficients that are located at the signal's cycle frequencies, for the estimation of the noise covariance, we need the coefficients lying between the cycle frequencies, which only carry estimation noise.
