ARH 312 Archaeological Laboratory: Frequently Asked Questions

ARH 312Y
Updated 22 March 2007

Your Questions: Radiocarbon

Course

Schedule

Evaluation

14C Lab

Reviews

ARH Prog

Anthro

Events

Rationale You will recall that I will not usually reply to your emails personally, unless they involve some question or problem particular to you. Instead, and especially if your question is relevant to other members of the class, I will post your question and my answer on this FAQ (Frequently Asked Questions) page.

Here are questions regarding the Radiocarbon Lab

Lab Questions
Question: My project is taking forever to calibrate. What's the problem?

Do NOT increase the checkpoint value from the default of 50000 iterations. If some students set this at millions of iterations, it ties up Sheffield's computer for days. You might even consider lowering the checkpoint value somewhat to speed things up .

Question: Under"Analysis and Interpretation" in the instructions, you ask us to find the shortest elapsed time that falls within the confidence interval and the likelihood that the period was that short or shorter. I'm not sure how to find these.

To get the date ranges for particular confidence levels, select "HPD Region" for the Elapsed time interval. Then you can type in any confidence level you want. Keep in mind that this region is similar to the confidence region withing a normal distribution when you do something like a Z-test (it just has a much more irregular shape). So there is a left tail (possible elapsed time shorter than the shortest within your confidence level), and a right tail (possible durations longer than your highest one within your confidence interval).

Question: For the lab assignment on BCal, I keep getting a "convergence checking error" when I submit my model. What should I do?

Answer: Generally you get the "convergence checking" error when your model is too complicated, too big, or contains some unresolvable contradiction.

The first thing to try is to reduce the convergence sensitivity (when you submit your model for calibration) from "5" to "4" or even "3."

If that doesn't work, make some copies of your project, and then edit each one to simplify it. For example, try reducing the number of groups by making two or three different models that overlap in time (basically, breaking up your model into parts). View the project and see if there are any dates that are major outliers. If so, eliminate them or at least assign them a high probability of being outliers.

If it still doesn't work, you probably have some impossible contradictions in your model. View the model and check the determinations listed for errors (like typos in the dates or incorrect stratigraphic orders). If there are impossible dates in your groups, or you put the groups in the wrong order, the model won't work.


ARH 312Y		Updated 22 March 2007
Your Questions: Radiocarbon		Updated 22 March 2007

Rationale		You will recall that I will not usually reply to your emails personally, unless they involve some question or problem particular to you. Instead, and especially if your question is relevant to other members of the class, I will post your question and my answer on this FAQ (Frequently Asked Questions) page.

		Here are questions regarding the Radiocarbon Lab
Lab Questions		Question: My project is taking forever to calibrate. What's the problem? Do NOT increase the checkpoint value from the default of 50000 iterations. If some students set this at millions of iterations, it ties up Sheffield's computer for days. You might even consider lowering the checkpoint value somewhat to speed things up . Question: Under"Analysis and Interpretation" in the instructions, you ask us to find the shortest elapsed time that falls within the confidence interval and the likelihood that the period was that short or shorter. I'm not sure how to find these. To get the date ranges for particular confidence levels, select "HPD Region" for the Elapsed time interval. Then you can type in any confidence level you want. Keep in mind that this region is similar to the confidence region withing a normal distribution when you do something like a Z-test (it just has a much more irregular shape). So there is a left tail (possible elapsed time shorter than the shortest within your confidence level), and a right tail (possible durations longer than your highest one within your confidence interval). Question: For the lab assignment on BCal, I keep getting a "convergence checking error" when I submit my model. What should I do? Answer: Generally you get the "convergence checking" error when your model is too complicated, too big, or contains some unresolvable contradiction. The first thing to try is to reduce the convergence sensitivity (when you submit your model for calibration) from "5" to "4" or even "3." If that doesn't work, make some copies of your project, and then edit each one to simplify it. For example, try reducing the number of groups by making two or three different models that overlap in time (basically, breaking up your model into parts). View the project and see if there are any dates that are major outliers. If so, eliminate them or at least assign them a high probability of being outliers. If it still doesn't work, you probably have some impossible contradictions in your model. View the model and check the determinations listed for errors (like typos in the dates or incorrect stratigraphic orders). If there are impossible dates in your groups, or you put the groups in the wrong order, the model won't work.