Guest Post: Analyzing the results of the Patent Agent examination
by Rajiv K. Choudhry
This post was prompted by the comments received on the post for results of the patent agent examination 2010.
A few words about the data analysis in the accompanying chart:
The patent office stated that the number of people who appeared for the exam was 1019. However, the data when completely collated showed 1018 people who took the exam (1274 registered and 256 absent). However, the figure taken by the patent office is assumed to be correct.
The pass percentages have been provided for each city with and without the absentees included. An average of marks, a standard deviation measure, and maximum and minimum marks in each part of the exam have also been included.
A measure of standard deviation (stddev) was also performed in order to see how much variation there is from the average scores. A low standard deviation indicates that the data points tend to be very close to the mean, whereas high standard deviation indicates that the data are spread out over a large range of values. As such, low standard deviation would indicate uniformity and a large standard deviation would indicate a disparity.
The average score for paper 1 and paper 2 in all 4 centers is remarkably similar-around 51 and 45 respectively. This may reflect to the objectivity of the question papers and a similar understanding of the questions amongst test takers. This is also evidenced from the standard deviation. The stddev for scores in Paper-1 and Paper-2 in Delhi, Chennai and Mumbai is very similar-14.5.
The standard deviation for viva is entirely a differnt matter for Mumabi, Chennai and Delhi. The standard deviation for viva should also be similar to that of paper-1 and paper-2. Statiscally, this is an anomaly and might reflect the subjectivity of the viva portion of the exam. The case of Kolkata highlights this portion-the stddev for viva and paper-1 and paper-2 is similar.
The parameter stddev is even more remarkable when considered for individual cases: it is seen that even though a candidates personal score in viva is the highest, they have failed overall (low marks in paper-1 and paper-2).
Another reason why the stddev is important is to see what is a more accurate reflection on the final result-paper-1/2 marks or the marks in viva. The average marks in Delhi for the passing candidate in paper 1, paper 2 and viva are 64, 62 and 70 respectively. However, for the candidates who failed in Delhi, the average marks in paper 1, paper 2 and viva are 47, 44, and 54 respectively. There is a strong co-relation between paper-1, paper 2 marks and pass and fail but there is a weak/very weak correlation between the viva marks and the overall result.
A complete gender based analysis was not performed but a small sample of the passing candidates in Delhi indicates that every 3 of the males who passed, 4 of the fairer sex qualified. To re-iterate, this a result from the small sample and not the overall data. This may also be biased because of assigning male and female name categories while seeing the results by the analyst. This result is nothing new: females out-perform males in most competetive examinations in the country. :-).
The gender of the Examiner is immaterial because the end result (pass/fail) is dependent upon paper-1 and paper-2 marks. An Examiner may boost the final result of the candidate by a few marks but that is the nature of an interview. Overall viva is not the deciding factor.
It is my personal opinion that because a major part of the communication between a patent agent and the patent office is written, the exam should also be a written exam where the patent office procedures are tested. This would be a true 'objective' exam.