韩宝成教授_Correlation.ppt

1. Measuring relationships Normally the variable study time correlate with the variable examination performance. Positive correlation: A score on one variable increases so the corresponding score on the other variable does the same. Negative correlation: a correlation between two variables where as one goes up the other goes down. Zero correlation: There is no relationship between two variables. There appears to be a positive correlation when we look at these results by eye but a clearer way to show this is to produce a scatter plot（散点图）. The points are not randomly scattered about the graph but generally fall within a band（带状）, indicating a correlation. But for random errors, the scores would have fallen along a line, the regression line（回归线）. If all the points lie exactly along a straight line then we have a perfect correlation（完全相关）. The measurement we use to describe the degree with which the points cluster along a straight line is the Pearson correlation coefficient（皮尔森相关系数）, r. Interpreting the correlation coefficient r=0, r2=0 Correlation between study time and exam performance Assumptions of the correlation coefficient The data should be interval data. Homoscedasticity（同方差性） (that is, if the intersection points between variables are plotted around the correlation ‘line’, they will be ‘normally’ distributed around the line – some will be above the line, some will be below it and more points will be close to the ‘line’ than far away). The relationship between the variables in question can be adequately portrayed by straight lines. Dekyser’s Study (2000): The Robustness of Critical Period Effects in SLA 2. Spearman’s Correlation Coefficient There will be times when we wish to correlate data that is not measured on a interval scale. The data would be measured on a ranked scale. we can perform a correlation on the ranks using the Spearman’s correlation coefficient. The Spearman coefficient is useful if we are concerned that the scores on two variables appea