This weekend, we ran the full regression on delta salary versus the number of citations. We observed that most professors received biyearly base salary raises, so we calculated the salary difference after two years and the corresponding citation changes. We were able to obtain 150 datapoints, and the result is as follows:
deltaSalary = beta * deltaCitations
| Estimate | p-value | |
|---|---|---|
| delta Citation | 33.4899 | 6.14027 x 10^-12 |
The p-value was extremely low, rejecting the null hypothesis.
Also, while running regression on biyearly data, we realized that we should have paired up the previous year's centrality with the current year's salary. We ran the regressions with centralities for 2009 with salaries for 2010, and here are the results:
| Estimate | p-value | |
|---|---|---|
| Constant | 57078.5 | 1.62923 x 10^-9 |
| Years Since PhD | 2431.5 | 2.01299 x 10^-10 |
| Citations | 5.062 | 0.00961352 |
| Estimate | p-value | |
|---|---|---|
| Constant | 60246.7 | 2.01867 x 10^-10 |
| Years Since PhD | 2338.52 | 9.00664 x 10^-10 |
| PageRank | 4.4146 x 10^6 | 0.0155467 |
The p-values and the estimates were similar to the previous results'.
During the last meeting, we agreed that we should try to capture the values of citations coming from differently ranked professors. This weekend, we debated whether PageRank would be a good measure to capture that. Even though the regression result we obtained for PageRank could be interpreted as statistically significant, we still haven't established the significance of delta PageRank.
Michael also wrote code for calculating g and h index. We will be analyzing the output this week.
No comments:
Post a Comment