I've just spent 16 hours on a lab. Never again.
December 13, 2009 6:30 PM Subscribe
What spreadsheet/graphing software should I use to do physics labs?
I am a high-school student, enrolled in the International Baccalaureate program. I'm taking an IB Physics class, and there's a lab component that's pretty substantial. What's worse, my physics teacher has his own very high expectations for labs, and trying to fulfill the standards of both is pretty difficult. Typically the labs are between 50-100 pages, about 60% graphs or charts. There is always too much data to do anything manually and maintain my sanity.
IB has a lot of bizarre niggling requirements for these labs. Every number has to have a unit and uncertainty associated with it (I'm not allowed to just put "+/- .5 V" at the top of a column) and significant figures are vitally important. Indeed, even a single error on one datum will drop me by the IB equivalent of a full letter grade, at least according to my teacher.
So, I would like some spreadsheet/graphing software that can do what I need automatically.
As far as numbers and calculations go:
-- there needs to be a way to associate units and uncertainties to numbers and still use them in calculations
-- calculate while propagating uncertainties
-- perform calculations using the proper numbers of significant figures
-- easily export charts, with units and uncertainties printed next to each number, for insertion into a word-processing document
As far as graphing goes:
-- It needs to be able to handle error-bars on graphs, and perform at least linear regression.
Does anything like this exist? I've looked around for formulas to make Excel work the way I want it to, but most of them are unsatisfactory. I've also tried out R, but I don't have time to program all this myself, and R seems pretty difficult to learn anyway.
Plan B is amphetamines
I am a high-school student, enrolled in the International Baccalaureate program. I'm taking an IB Physics class, and there's a lab component that's pretty substantial. What's worse, my physics teacher has his own very high expectations for labs, and trying to fulfill the standards of both is pretty difficult. Typically the labs are between 50-100 pages, about 60% graphs or charts. There is always too much data to do anything manually and maintain my sanity.
IB has a lot of bizarre niggling requirements for these labs. Every number has to have a unit and uncertainty associated with it (I'm not allowed to just put "+/- .5 V" at the top of a column) and significant figures are vitally important. Indeed, even a single error on one datum will drop me by the IB equivalent of a full letter grade, at least according to my teacher.
So, I would like some spreadsheet/graphing software that can do what I need automatically.
As far as numbers and calculations go:
-- there needs to be a way to associate units and uncertainties to numbers and still use them in calculations
-- calculate while propagating uncertainties
-- perform calculations using the proper numbers of significant figures
-- easily export charts, with units and uncertainties printed next to each number, for insertion into a word-processing document
As far as graphing goes:
-- It needs to be able to handle error-bars on graphs, and perform at least linear regression.
Does anything like this exist? I've looked around for formulas to make Excel work the way I want it to, but most of them are unsatisfactory. I've also tried out R, but I don't have time to program all this myself, and R seems pretty difficult to learn anyway.
Plan B is amphetamines
Origin is a fairly standard package for this purpose. The plots look pretty good, and it handles the error bars easily. Highly recommended over Excel. I'd only use excel for plotting as a last resort. It will make pretty tables though, so it's not all bad.
Also, nothing is going to propagate errors for you, nor use significant figures. This is something you have to do by hand. But if you get good and familiar with error propagation, you'll be a mile ahead of everyone else in college-level intro labs.
posted by kiltedtaco at 6:41 PM on December 13, 2009
Also, nothing is going to propagate errors for you, nor use significant figures. This is something you have to do by hand. But if you get good and familiar with error propagation, you'll be a mile ahead of everyone else in college-level intro labs.
posted by kiltedtaco at 6:41 PM on December 13, 2009
I prefer MATLAB to Origin, but either option is hands-down more favourable to Excel IMHO.
posted by Cody's Keeper at 6:51 PM on December 13, 2009
posted by Cody's Keeper at 6:51 PM on December 13, 2009
Igor is the software used in labs at my college. It has demo versions available.
When I was in high school, we used Vernier LabQuest software for IB science classes; the school had a site license, and the teacher could legally burn each student a copy for use at home. You might ask your teacher about something like that. As I recall, the software was pretty good -- it did error bars very nicely, in particular.
I, personally, use MATLAB for most of my graphing needs, but I get my copy through the college; I have no idea how a high school student might get a license.
posted by Commander Rachek at 7:09 PM on December 13, 2009
When I was in high school, we used Vernier LabQuest software for IB science classes; the school had a site license, and the teacher could legally burn each student a copy for use at home. You might ask your teacher about something like that. As I recall, the software was pretty good -- it did error bars very nicely, in particular.
I, personally, use MATLAB for most of my graphing needs, but I get my copy through the college; I have no idea how a high school student might get a license.
posted by Commander Rachek at 7:09 PM on December 13, 2009
Just came here to say that I took IB Physics HL, and that there's no way your labs should be 50-100 pages long. My longest lab was 9 pages, and I received a 7 in the end. 100 pages is just insane, no matter how many figures. I don't think I ever used anything other than Excel at the time, which was always more than enough.
posted by spaghettification at 7:13 PM on December 13, 2009
posted by spaghettification at 7:13 PM on December 13, 2009
Response by poster: Just came here to say that I took IB Physics HL, and that there's no way your labs should be 50-100 pages long. My longest lab was 9 pages, and I received a 7 in the end. 100 pages is just insane, no matter how many figures. I don't think I ever used anything other than Excel at the time, which was always more than enough.
Yes, I realize this. Unfortunately, my physics teacher is even crazier than a physics teacher has a right to be, and it's what he expects. (I think it's a paranoid response to one year, a long time ago, when most of his students received 1s and 2s on the lab portion.) To be fair to him, the school did get a note back from the IBO saying they were astounded by the quality of our physics labs.
As for MATLAB/Origin, they both offer student licenses for relatively cheap. Would these honestly solve my problems, or are they just better than Excel? My needs aren't complicated at all apart from the significant figure and unit issues. Also, I have access to IDL because of an internship. Could it be helpful in the same way as these packages?
posted by vogon_poet at 7:21 PM on December 13, 2009
Yes, I realize this. Unfortunately, my physics teacher is even crazier than a physics teacher has a right to be, and it's what he expects. (I think it's a paranoid response to one year, a long time ago, when most of his students received 1s and 2s on the lab portion.) To be fair to him, the school did get a note back from the IBO saying they were astounded by the quality of our physics labs.
As for MATLAB/Origin, they both offer student licenses for relatively cheap. Would these honestly solve my problems, or are they just better than Excel? My needs aren't complicated at all apart from the significant figure and unit issues. Also, I have access to IDL because of an internship. Could it be helpful in the same way as these packages?
posted by vogon_poet at 7:21 PM on December 13, 2009
MATLAB isn't actually that easy to work with for these types of graphs; it's very versatile, but there's a steep learning curve. I use it for most of my graphing in large part because I'm a math major, and most of my graphs are of data generated by MATLAB scripts. If you're willing to put in the time, knowing how to use it is a valuable skill, but you'll have to put in a lot of time. I'd recommend using Igor or LabQuest over MATLAB.
Your physics teacher is crazy; I think he is crazier than you quite realize. You should not be spending 16 hours on lab reports, and they should not be more than 10 pages long at the most. To give you some perspective: the typical undergraduate honors thesis in my physics department is about 50-100 pages long and encompasses a year or more of work. I highly doubt you are spending a year on each of your labs.
If this is severely impacting your quality of life and/or other school work (which I imagine it is), you and your parents should complain to the administration, preferably in the company of other students and their parents. You'd be shooting yourself in the foot to put up with this.
posted by Commander Rachek at 7:44 PM on December 13, 2009
Your physics teacher is crazy; I think he is crazier than you quite realize. You should not be spending 16 hours on lab reports, and they should not be more than 10 pages long at the most. To give you some perspective: the typical undergraduate honors thesis in my physics department is about 50-100 pages long and encompasses a year or more of work. I highly doubt you are spending a year on each of your labs.
If this is severely impacting your quality of life and/or other school work (which I imagine it is), you and your parents should complain to the administration, preferably in the company of other students and their parents. You'd be shooting yourself in the foot to put up with this.
posted by Commander Rachek at 7:44 PM on December 13, 2009
Also, correction: the software we used in high school was Logger Pro, also by Vernier; LabQuest is the software that runs on the hand-held data collection devices.
posted by Commander Rachek at 7:55 PM on December 13, 2009
posted by Commander Rachek at 7:55 PM on December 13, 2009
I don't even remember the lab portion of my IB Physics exam. I had one of the most burnt out teachers ever (we all loved him regardless), and I am sure we had to put together some lab work in lab notebooks, but ours were actually all hand-written in comp books, I think. I got a 5 or 6 on the exam. Even my college physics lab (electronics!) only had us do hand-written labs. Honestly, I think labs are easiest to write out as you're going along, graphs included. How are you collecting the data as you go?
Honestly, I'd try asking the teacher what he suggests. Tell him that things are taking forever, and if there's anything he suggests for keeping it easy. Does he realize that the labs are taking you forever? Is it just you? Is it everyone?
posted by that girl at 8:23 PM on December 13, 2009
Honestly, I'd try asking the teacher what he suggests. Tell him that things are taking forever, and if there's anything he suggests for keeping it easy. Does he realize that the labs are taking you forever? Is it just you? Is it everyone?
posted by that girl at 8:23 PM on December 13, 2009
The data analysis software I used for this kind of thing in college and grad school was KaleidaGraph. It will do pretty much everything you've asked for here (although see kiltedtaco's comments about propagating uncertainties.)
posted by Johnny Assay at 8:43 PM on December 13, 2009 [1 favorite]
posted by Johnny Assay at 8:43 PM on December 13, 2009 [1 favorite]
Kiltedtaco is right: you'll have to set up the error formulas yourself. But after that, Excel should be able to generate the chart you want. I've found it can usually do what I want it to, chart-wise, although the answer is often not obvious and sometimes involves code. The best online resource I know of is http://peltiertech.com/Excel/Charts/ChartIndex.html
I've also used http://www.mrexcel.com/
As far as automatically generating scientific charts, I think R is in some ways the best for that; there are many sophisticated chart options. However, as you observed, it would take a lot to learn it. I think the most time-effective approach is excel. Do your work in advance, identify the few things you need to add, hone your google searching skills, and when all else fails, post a message asking for help and get some sleep. I don't know the excel google group, but I wouldn't be surprised if there's depth there. There are a lot of people who have mad Excel chops.
Good luck!
posted by manduca at 8:51 PM on December 13, 2009
I've also used http://www.mrexcel.com/
As far as automatically generating scientific charts, I think R is in some ways the best for that; there are many sophisticated chart options. However, as you observed, it would take a lot to learn it. I think the most time-effective approach is excel. Do your work in advance, identify the few things you need to add, hone your google searching skills, and when all else fails, post a message asking for help and get some sleep. I don't know the excel google group, but I wouldn't be surprised if there's depth there. There are a lot of people who have mad Excel chops.
Good luck!
posted by manduca at 8:51 PM on December 13, 2009
From my university days (in the last century) I remember using Origin, mentioned above, and Graphpad Prism, which handled this kind of stuff brilliantly. Those were the win 3.1 versions (I also remember using an Inplot-something which ran under DOS), but I suppose not much has changed (not in the underlying math!)
The way I see it, Excel is a financial analysis spreadsheet software; it can do data analysis, linear regression, plots, etc., but much in the same way you can use the heel of your shoe to drive a nail. A hammer is way better.
posted by _dario at 9:10 PM on December 13, 2009
The way I see it, Excel is a financial analysis spreadsheet software; it can do data analysis, linear regression, plots, etc., but much in the same way you can use the heel of your shoe to drive a nail. A hammer is way better.
posted by _dario at 9:10 PM on December 13, 2009
Not the answer to your question, but if I were a student in physics I would like this program a lot: (it is free)
http://ccsl.mae.cornell.edu/eureqa
What it is in a nutshell: It is a spreadsheet like thing that you can enter raw data into and using AI it can figure out what equation makes the numbers relate.
awesome!
There is a two part easy to understand video on the site as well that explains it better than I just did.
posted by santogold at 11:02 PM on December 13, 2009 [2 favorites]
http://ccsl.mae.cornell.edu/eureqa
What it is in a nutshell: It is a spreadsheet like thing that you can enter raw data into and using AI it can figure out what equation makes the numbers relate.
awesome!
There is a two part easy to understand video on the site as well that explains it better than I just did.
posted by santogold at 11:02 PM on December 13, 2009 [2 favorites]
Assuming you're reading your data from either CSV or tab-delimited files, these can be handled easily in MATLAB. Then you just program in your formulas and can perform the calculations in a script. This is nice, because you can automate repetetive tasks, and just copy and paste snippets of code that you use. If you're at all comfortable programming, or able to learn it quickly, it should come naturally for you. If programming is something you find difficult, MATLAB is probably not as good an option.
Tracking units doesn't seem like it should be that hard. You should just have two units to deal with per graph, right? (x and y axes).
Uncertainty propagation is a bitch, but you basically just have to find the formulas, then program them in, and then let the software calculate it.
posted by !Jim at 11:17 PM on December 13, 2009
Tracking units doesn't seem like it should be that hard. You should just have two units to deal with per graph, right? (x and y axes).
Uncertainty propagation is a bitch, but you basically just have to find the formulas, then program them in, and then let the software calculate it.
posted by !Jim at 11:17 PM on December 13, 2009
R is a powerful (and free) statistical computing and plotting environment and programming language. It also makes awesome plots that you can export to, e.g., .png, .svg, .ps, or .pdf, and is great for things like regression.
There is a bit of a learning curve, though, since the syntax can be idiosyncratic. If you end up using it I would recommend checking out a book on R or S+ (they're basically the same) from the local library as a reference. You can also try googling through the R mailing list, since for most questions someone has probably asked something similar (and gotten an accurate but terse reply).
posted by en forme de poire at 7:44 AM on December 14, 2009
There is a bit of a learning curve, though, since the syntax can be idiosyncratic. If you end up using it I would recommend checking out a book on R or S+ (they're basically the same) from the local library as a reference. You can also try googling through the R mailing list, since for most questions someone has probably asked something similar (and gotten an accurate but terse reply).
posted by en forme de poire at 7:44 AM on December 14, 2009
Excel is a very difficult tool to do this in, in my experience. A quick example: Imagine a relative uncertainty of 2% with two measurements: 12.256 (+/- 0.245) and 1.2253 (+/- 0.0245). In proper notation, this would be written 12.3 +/- 0.2 and 1.22 +/- 0.02. Note that the number of decimal places are different for each result and that the rounding rules are different for each number (here using the 5+even round up and 5+odd round down convention---I don't know what the IB uses). Excel can do this with a lot of hand-holding, but will get it wrong by default. As a lab instructor, I failed many a student over these errors. Excel has very perverse defaults for standard scientific tabulation.
In my experience, there's really no alternative for tables to type-set them yourself, either in Word or LaTeX (preferred). R can come pretty close, in fact, exactly reproduce your requirements, but will require a fair bit of set up.
For graphs, I personally use SigmaPlot, with which it's fairly easy to produce exact camera-ready copy. Origin is equivalent. R is amazing at producing graphs, but like LaTeX, will require a significant time investment. Excel can make graphs, simple ones at least, but only with enormous effort and tweaking (again, the defaults are all against you here).
Finally, another word of warning: Excel uses a non-standard (i.e. wrong) math library, optimized for speed rather than correctness. Thus, regressions are problematic in excel, except for maybe linear ones, due to weird, unreproducible arithmetic errors. R or Matlab or some other scientific-math/stats package will not have this problem.
posted by bonehead at 9:41 AM on December 14, 2009
In my experience, there's really no alternative for tables to type-set them yourself, either in Word or LaTeX (preferred). R can come pretty close, in fact, exactly reproduce your requirements, but will require a fair bit of set up.
For graphs, I personally use SigmaPlot, with which it's fairly easy to produce exact camera-ready copy. Origin is equivalent. R is amazing at producing graphs, but like LaTeX, will require a significant time investment. Excel can make graphs, simple ones at least, but only with enormous effort and tweaking (again, the defaults are all against you here).
Finally, another word of warning: Excel uses a non-standard (i.e. wrong) math library, optimized for speed rather than correctness. Thus, regressions are problematic in excel, except for maybe linear ones, due to weird, unreproducible arithmetic errors. R or Matlab or some other scientific-math/stats package will not have this problem.
posted by bonehead at 9:41 AM on December 14, 2009
perform calculations using the proper numbers of significant figures
BTW, I certain that you know this, but this isn't desirable in the least. You want to do calculations at i) infinite precision, and, in the real world, at high-precision (lots of decimal places). The rule of thumb we used for numerical computations was minimum 6 (decimal) digits of precision more than significant, more if we could get them. You also want to do them with reproducible precision, i.e., to some standard like IEEE 754. Sigfigs rules are applied to the computed result, based on the input (experimental) relative values.
This all boils down to: Don't round-off intermediate results. It can really mess up your answers.
posted by bonehead at 9:48 AM on December 14, 2009
BTW, I certain that you know this, but this isn't desirable in the least. You want to do calculations at i) infinite precision, and, in the real world, at high-precision (lots of decimal places). The rule of thumb we used for numerical computations was minimum 6 (decimal) digits of precision more than significant, more if we could get them. You also want to do them with reproducible precision, i.e., to some standard like IEEE 754. Sigfigs rules are applied to the computed result, based on the input (experimental) relative values.
This all boils down to: Don't round-off intermediate results. It can really mess up your answers.
posted by bonehead at 9:48 AM on December 14, 2009
This thread is closed to new comments.
posted by dfriedman at 6:32 PM on December 13, 2009