Is there a score for Panda, like there is for Pagerank, or are you either in the panda penalty or not?
I love this question. Not everything is a going to be boolean of course. For example, domain age cannot be boolean by itself. The simplest form of AI is building decision trees and decision tables. Building decision trees are generally boolean though not always. I will explain briefly how binary decision trees and tables are used most of the time for things like this. Then I will try and answer the question.
Please note: I am not saying this is exactly how SEO works, but you can use this as an illustration of often how these things work.
First things first. You have to create a list of questions based upon the criteria that you want to measure. For example, Are there links to external sites in the footer? and, Are these links site-wide? Any question that is not a yes or no question is whittled down to criteria to answer that particular question. Each question and criteria are simple yes and no questions and written so that yes is good and no is bad. Binary 1 for yes, 0 for no. The list of questions can be rather long of course, but that does not matter. As you can imagine, some questions do not have to be asked if the answer to a particular question is a yes or no. For example, I do not need to know if external links in the footer are site-wide if there are no external links in the footer. This is where the questions are organized in a hierarchical structure based upon dependency. It is a simple flow-chart. Not all questions have to have an answer and the flow chart is used to determine what questions are asked and which are not and when a question is asked. This is a decision tree.
Then each question is used to evaluated conditions- in this case SEO factors- with a binary yes or no. Code is written to evaluate the answer to each question as either being true or false. A simple string is concatenated with the binary 1 or 0 for each of the measured criteria. It would be something like "1110100110101" but much longer. Simple binary math is used to determine a decimal value. Some criteria are show stoppers. A simple bit evaluation for the binary value for the criteria is rather easy to do. For example, is bit 43 a 1 or 0? This can be done, but often not. Instead, a table of values and/or value ranges for the decimal number is calculated and and kept for reference. If the number is greater than x, equal to x, or between x and y, then = (equals) decision. This can determine hard conditions as well as soft conditions. For example, if the number is between 43021 and 43057, this is a penalty condition. The right bits are true to end up with a number in this range.
Okay. So I have over simplified the process, but you get the point. For the most part, for SEO with a few exceptions such as domain age, think of SEO mostly as what a site needs to do for performance in a binary way. Yes I have links to external sites site-wide in the footer. Remove them.
Is Panda binary or is there a score? Google keeps a few scores. SiteRank, TrustRank, and PageRank. SiteRank is the old concept of PageRank as it was originally used- site wide. TrustRank is a part of SiteRank. And PageRank, as it stands, is calculated on a page by page basis. You can imagine that a penalty cannot be derived from this. A decision table of some sort is likely used. And since criteria equals a number that is used against a decision table, you can call it a score. But it is not really. Certainly, how Google makes decisions is not shared. So no. No score. Not really. But know that the criteria score will effect TrustRank thus SiteRank. It will be simple math. Perhaps one of the reasons why Google does not want to maintain PageRank site-wide anymore.
So to answer your question, Panda and other penalties are based upon simple binary criteria that a site meets. How that criteria is divided up is anyone's guess, but know that in the end you can boil it down to simple things that a site has done that Google does not like. Reversing or removing these simple things that Google does not like will in time clear the condition within Google's eyes and lift the penalty. Think of SEO in binary terms, and you will be far ahead of the rest of the class.
[Update] JMC made a great point and I added a comment that should live on. So here it is.
We are conditioned to think of Panda and others as a single entity. It is likely not. It is a deployment of several updates to algorithms. Knock out one violation, performance improves. Knock out another, more improvement. This tells me that Panda is not an entity, but rather a collection. If you remember the decision table, each violation likely adds a down-bump in the scores that determine SERP placement. If you fix one condition, that down-bump disappears hence a gain in performance. Knock them all out, then no down-bump at all. While I suspect that the bulk of Panda algorithms apply site wide, it may not be all site-wide but also contain changes that effect single pages too.
And that is what we must remember in SEO. Penalties do not come easily. Just stay out of the way of the Google Quartermaster, and you should be fine. Pay attention to what Google says you should not do and take some of their advice with salt- after all, it is to benefit Google more than it is to benefit you. I have seen Google say some terribly silly things and reference things that do not work. But do pay attention to the no no(s)!
Personal experience with client sites makes me believe you can shed Panda in chunks. As in, currently site X is experiencing a 75% panda penalty. After some work and the next refresh it is now experiencing a 50% penalty indicated by traffic increases at the same time as a Panda refresh. It should be noted that other factors could be at work in this case and the traffic increase has nothing to do with Panda's affect on your site, but instead in how it affected others.
However when a Panda refresh happens and a site suddenly triples its traffic in a short period of time over all pages (which i've seen), it makes it seem like Panda is a matter of penalty or no penalty. The key signal of a reduced or removed penalty is that the increase is dramatic and across the entire site, even on pages that are lower quality compared to higher quality pages on the site.
To summarize, I think you can incrementally remove Panda from certain Pages on a site, while the entire site continues to suffer an overall Panda penalty.