Check your assumptions for linear regression. Are both variables continuous random variables? How do you know? Plot the two variables using a scatterplot. Are the variables linearly related? If the variables are not linearly related, try transforming one or more variables and generate another scatterplot with your transformed variable(s) until you are satisfied with the linear relationship. Plot a residuals versus fits plot and explain whether or not the plot shows errors that are random and independent, errors have a mean of 0, and errors have the same variance. Perform a normality plot of the residuals and check the Ryan-Joiner statistical test for normality. If the RJ statistic is > 0.05, then you reject the null that the residuals are not normal and accept the alternate hypothesis that residuals are normally distributed. So, for normality to apply, you want to fail to reject the null (large p-value).