(SILVA, Glauco P. GUARNIERI, Fernando H. Comments on is Statistical Significance not Significant? Brazilian Political Science Review. Vol.8, No 02, 2014)It is very rewarding for us to receive serious commentary on When is Statistical Significance not Significant?. We are pleased with Silva and Guarnieri's (2014) remarks and we believe that they generally agree with us. However, their review makes it clear that some points were left behind. The principal aim of this paper is to answer Silva and Guarnieri's (2014) comments on Figueiredo Filho et al. (2013). Methodologically, we use both observational and simulation data to defend our view on the proper use of the p-value statistic in empirical research.(1) Scholars must always graphically analyze their data before interpreting the p-valueIn many cases, as pointed out by Silva and Guarnieri (2014), graphical cannot help you. That being said, ignoring graphs is much worse path to trail. Graphical is powerful tool not only for examining linear relationships but also to identify exponential, quadratic, and cubic relationships.Additionally, graphical can be applied to more descriptive goals not related to the presence of covariates or model selection. We simulated an independent t test comparison between the heights of men and women. For both groups the distribution is normal. Men have an average of 1.75m with standard deviation of .15. Women have an average of 1.60m with standard deviation of .10. Figure 1 illustrates the data.When there is no outlier, as long that there is no overlap between confidence intervals, we may conclude that men are taller than women in the population. The mean difference between groups is statistically significant (p-valueIn some specific areas of Statistics, graphs are fundamental step of the scientific initiative. The selection of the appropriate specification in time series depends heavily on graphs. Let us examine data from Box and Jenkins (1976)1.We observe strong seasonality, tendency and increasing variance over time. We must graphically examine the original distribution of the variables before choosing the appropriate model.Using both graphical and adjustment measures, we define the model order that best fits the data. In this case SARIMA (0,1,1) (0,1,1). Graphical is at the heart of all statistical analysis.Now let us deal directly with Silva and Guarnieri's (2014) example regarding Taagepera's (2012) experiment. They argue that a simulation of this data shows that the graphical evaluation would not be enough to avoid misguided analysis (SILVA AND GUARNIERI, 2014, p. 02). We disagree. To make our case we simulated table of values of y, x1, x2 and x3 following Taagepera (2012). All values are random and the y value came from = 980x1x3/x22 . The next step is to fit linear model to explain the variance of y and graphically analyze the residuals. Figure 4 displays this information.Graphical examination of the standardized residuals and predicted values shows that the relationship is not linear. Graphical reveals that the linear function is not appropriate to model y. We should never adjust regression models without relying on residuals inspection. Silva and Guarnieri (2014) also argue that theory should inform the adequate functional form. We completely agree with them on this. However, sometimes data defies theoretical expectations and at times we do not have strong theoretical assumptions to follow. In the total absence of theoretical guidance, graphical can help scholars in more inductive pattern. …