Scaling Woes!

Benjamin_Gates

Hi Everyone

I was wondering (as discussed in some other posts) if UPSC could deliberately take some optional for a ride (say Public Administration) in the process called Scaling, or the process is applied to all the optionals uniformly. This year many aspirants (including me) got double digit marks in paper 1 of Pub Ad. Same was the case of Anthropology. I wrote Mains in 2011 and 2012, casual attempts, in 2011, I got what I thought I would - 200+ in each of these two subjects but the reults in 2012 were totally unexpected especially paper 1 of both - I deserved more. Also, is there some sort of "trend" (up/downscaling any optional for a number of years) involved here?

Benjamin_Gates

To simplify:

Suppose if 60 % of selected candidates have one optional in common, what UPSC is going to do about it ? Stricter marking? Tougher Questions? Additional downscaling? or just ignore the stats altogether?

I've read and people have told me that scaling process brings all optionals' marks at one level. I was wondering if this "Additional" downscaling exists for (more popular) optionals like Pulic Administration or not

Benjamin_Gates

Sorry, I meant Public Administration in the last line.

Stardust

To simplify:

Suppose if 60 % of selected candidates have one optional in common, what UPSC is going to do about it ? Stricter marking? Tougher Questions? Additional downscaling? or just ignore the stats altogether?

I've read and people have told me that scaling process brings all optionals' marks at one level. I was wondering if this "Additional" downscaling exists for (more popular) optionals like Pulic Administration or not

There really is no discernible logic in the scaling. If something happened one year does not mean it will happen the next year. Its best to ignore all this, do your best and hope for some luck.

Benjamin_Gates

@Stardust1901 thanks a lot for the reply
Perhaps this thread was created out of anxiety (no sleep for 28 hours now) and has no scope for meaningful discussion. I request @moderators to close it.

skylax

There is a thread on this forum that quotes from UPSC's documents and explains the scaling process. Search it (it was posted early this year), read it in leisure and figure out your answers.

As I understand it, regression would hit the the subject with lot of people and lower scores harder than subjects with less number of people and higher scores.

Another takeaway from the method is that the final score (in optional subjects) is definitely not the one on one's answer sheets as awarded by the examiner, barring a statistical miracle. Because they will get up-scaled or down-scaled by regression. People were wondering in the forum, after the marks were out two months back, as to how they could get so-and-so marks after they had attempted all the questions (in Pub Ad especially). I think scaling is the answer.

Benjamin_Gates

There is a thread on this forum that quotes from UPSC's documents and explains the scaling process. Search it (it was posted early this year), read it in leisure and figure out your answers.

As I understand it, regression would hit the the subject with lot of people and lower scores harder than subjects with less number of people and higher scores.

Another takeaway from the method is that the final score (in optional subjects) is definitely not the one on one's answer sheets as awarded by the examiner, barring a statistical miracle. Because they will get up-scaled or down-scaled by regression. People were wondering in the forum, after the marks were out two months back, as to how they could get so-and-so marks after they had attempted all the questions (in Pub Ad especially). I think scaling is the answer.

@skylax thanks a lot for the reply

This one?

http://forumias.com/discussion/728/scaling-in-upsc-civil-services-mains-examination-explained/p1

I read that. Your point on no. of students and higher/lower scores was perhaps what I was thinking over again and again. (w.r.t. to Anthropology & Pub Ad). At least the scaling will have less effect due to just 1 optional this time around.

skylax

Yes, that thread. And yes, we can find solace in the knowledge that Essay and GS which would make up about 70% of written won't have scaling (as against 40% in previous years) and that the subjects that generally benefit from scaling (not necessarily literature, but any subject with very few data points i.e. no. of examinees) won't be that helpful in GS and Essay papers (and interview to an extent) unlike Pub Ad, Anthropology, Political Science etc.