Internet Archaeol.13 Christen. Software Usage

<BODY>  <SCRIPT SRC="/journal.js"></SCRIPT> <hr><p style="text-align:center"> <a href="1intro.html" onMouseOver="on('backt');window.status='Go Back';return true" onMouseOut="off('backt');window.status='';return true"> <img src="/icons/back.gif" name="backt" ALT="[Back]" border="0"></a> <a href="3int.html" onMouseOver="on('fwdt');window.status='Go Forward';return true" onMouseOut="off('fwdt');window.status='';return true"> <img src="/icons/fwd.gif" name="fwdt" ALT="[Forward]" border="0"></a> <a href="toc.html" onMouseOver="on('toct');window.status='Go to Article Table of Contents';return true" onMouseOut="off('toct');window.status='';return true"> <img src="/icons/toc.gif" name="toct" ALT="[Contents]" border="0"></a> <a href="/" onMouseOver="on('homet');window.status='Go to Internet Archaeology Home Page';return true" onMouseOut="off('homet');window.status='';return true"> <img src="/icons/ia.gif" name="homet" ALT="[Journal Homepage]" border="0"></a> </p>  <H2>2 Bwigg software usage</H2> <P>Bwigg is an interface to the general purpose Bayesian calibration software mexcal. After formatting the input data, Bwigg collects the output from mexcal and calls the graphics software <a href="http://www.gnuplot.info">gnuplot</a> to create the figures. Bwigg then creates an HTML document with an outline of the results, with links for output file retrieval.</P> <P>An interface for mexcal, including several features for Bayesian calibration (not including wiggle matching, however), may be found in <a href="http://intarch.ac.uk/journal/issue7/buck_index.html">BCal</a>. Other software for radiocarbon calibration includes <a href="http://www.rlaha.ox.ac.uk/oxcal/oxcal.htm">OxCal</a>, <a href="http://www.cio.phys.rug.nl/HTML-docs/carbon14/cal25.html">Cal25</a> and <a href="http://www.calpal.de">CalPal</a>. Access Bwigg at <a href="http://www.cimat.mx/Bwigg">http://www.cimat.mx/Bwigg</a>. Reading all the text in the introductory page is <strong>strongly</strong> recommended.</P> <P>It is assumed that the set of radiocarbon determinations to be calibrated form a floating chronology. First, the determinations should be sorted starting with the latest (youngest) sample in the calendar scale. For tree-rings, this would be the outer-most sample in the sequence. Second, the event of interest should be established. For example, suppose the bark was not radiocarbon dated (say, 30 tree-rings until the border of the log), but the event to be dated is when the tree was felled. In this case the age for the first sample, θ<sub>1</sub>, is 30 years before the event to be dated. Therefore, using years <em>Before Present</em> (BP; taking as the 'present' the year AD 1950), should be</P> <div class="centred">θ<sub>1</sub> = θ<sub>0</sub> + 30 </div> <P>In such a case α<sub>1</sub> = 30 should be set. The same should be done for the rest of the determinations, to complete a list such as the one presented in Table 1 (although for the data set in Table 1, θ<sub>0</sub> = θ<sub>1</sub>, i.e. α<sub>1</sub> = 0). </P> <a name="tab1"></a><div class="centred"> <table border="1" cellpadding="5"> <tr bgcolor="#ccccff"><td>j</td><td>α<sub>j</sub></td><td>r.c det: y<sub>j</sub> ± σ<sub>j</sub></td><td>Lab. Number</td></tr> <tr><td>1</td><td> 0</td><td>2180 ± 25</td><td>4084</td></tr> <tr><td>2</td><td>20</td> <td>2187 ± 20</td><td>4083</td></tr> <tr><td>3</td><td>40</td> <td>2210 ± 25</td><td>4082</td></tr> <tr><td>4</td><td>60</td> <td>2225 ± 25</td><td>4081</td></tr> <tr><td>5</td><td>80</td> <td>2217 ± 20</td><td>4080</td></tr> <tr><td>6</td><td>100</td> <td>2247 ± 25</td><td>4079</td></tr> <tr><td>7</td><td>120</td> <td>2280 ± 25</td><td>4078</td></tr> <tr><td>8</td><td>140</td> <td>2317 ± 25</td><td>4077</td></tr> <tr><td>9</td><td>160</td> <td>2335 ± 20</td><td>4076</td></tr> <tr><td>10</td><td>180</td> <td>2358 ± 25</td><td>4075</td></tr> <tr><td>11</td><td>200</td> <td>2387 ± 30</td><td>4074</td></tr> <tr><td>12</td><td>220</td> <td>2405 ± 25</td><td>4073</td></tr> <tr><td>13</td><td>240</td> <td>2432 ± 30</td><td>4072</td></tr> <tr><td>14</td><td>260</td> <td>2447 ± 20</td><td>4071</td></tr> <tr><td>15</td><td>280</td> <td>2468 ± 25</td><td>4070</td></tr> <tr><td>16</td><td>300</td> <td>2493 ± 20</td><td>4069</td></tr> <tr><td>17</td><td>320</td> <td>2528 ± 30</td><td>4068</td></tr> </table> </div> <p class="image">Table 1: Radiocarbon determinations for the Ulandryk 4 kurgan, Altai Mountains, southern Siberia, dendrosample 19116 (see Slusarenko <i>et al</i>. <a href="biblio.html#slusarenko01">2001</a>). In this case θ<sub>0</sub> = θ<sub>1</sub> (α<sub>1</sub> = 0). The laboratory is the Radiocarbon Laboratory of the Institute of Geology, Siberian Branch of the Russian Academy of Sciences in Novosibirsk (standard liquid scintillation counting).</p> <P>Table 1 is taken from Slusarenko <i>et al</i>. (<a href="biblio.html#slusarenko01">2001</a>). A log found in a burial place in southern Siberia, belonging to the <em>Pazyryk</em> culture, was analysed. The tree rings were sub-divided into groups of 20 rings each, and seventeen samples (340 rings counted from the centre), were selected for dating, as shown in Table 1 (see details in Slusarenko <i>et al</i>. <a href="biblio.html#slusarenko01">2001</a>).</P> <P>The next step is to access the data input page of <a href="http://www.cimat.mx/Bwigg/wigg.html">Bwigg</a>. A name for the project (task) for Bwigg should be selected. It is recommended that a name characteristic of the dating study is selected. For example, for the data in Table 1 'Ulandryk' was chosen. Next, the data should be entered. The first line, and only the first line, is available (and should be used) for a comment. This line may be used for a short description of the data. Exceeding the width of the input window is not recommended. The data come next. Each line corresponds to one radiocarbon determination. The format for each line is as follows:</P> <div class="centred"> α<sub>j</sub>    y<sub>j</sub> ± σ<sub>j</sub> LabNumber </div> <P>The data may first be typed in a plain text editor (e.g. Notepad in Windows, for example, but not Word), and then copied and pasted into the input window. The data from Table 1, ready to be pasted into the input window look like this:</P> <pre> Ulandryk 4 kurgan, Altai Mountains, southern Siberia, dendrosample 19116. 0 2180 ± 25 4084 20 2187 ± 20 4083 40 2210 ± 25 4082 60 2225 ± 25 4081 80 2217 ± 20 4080 100 2247 ± 25 4079 120 2280 ± 25 4078 140 2317 ± 25 4077 160 2335 ± 20 4076 180 2358 ± 25 4075 200 2387 ± 30 4074 220 2405 ± 25 4073 240 2432 ± 30 4072 260 2447 ± 20 4071 280 2468 ± 25 4070 300 2493 ± 20 4069 320 2528 ± 30 4068 </pre> <P>The last step is to provide a period of time (early and late boundaries) for calibration. This constitutes the <em>a priori</em> distribution for θ<sub>0</sub> (see <a href="3int.html">Bayesian wiggle-matching</a>), which is uniform within the early and late boundaries. Although a very wide range may be used, it is strongly recommended that all available knowledge is incorporated, <em>not including the current radiocarbon determinations</em>, to give an educated guess for this period, thus making the range as short as possible (remember that Bwigg only uses years BP). The data are now sent with the <em>send</em> button. Note that a 5% <em>a priori</em> probability is used for each determination being an outlier; see <a href="3int.html#3.1">Outliers</a> for details.</P> <P>After a short processing time, the output report page shows up in your browser. First of all, go to section 4 of the report (Outlier analysis) and write down or copy some or all of the <em>posterior probabilities</em> given. Go back to the input page and re-send the data. The posterior probabilities in the refreshed calculations should be roughly equal to the first calculations. These probabilities are numerically approximated using a simulation method (called MCMC; see <a href="4tech.html#4.1">technical</a> section). Problems may arise if the time span (early and late boundaries) given to Bwigg is far greater than the span of the posterior distributions (Figures <a href="#fig1">1</a> and <a href="#fig2">2</a>). After some assurance that the approximations are correct, you may turn to analyse the results (note that the posterior distributions, Figures 1 and 2 in the Bwigg report, are calculated exactly).</P> <P>The results are given in the output page. Here the posterior distributions are reproduced in Figures <a href="#fig1">1</a> and <a href="#fig2">2</a>, not considering and considering the errors in the calibration curve respectively. Also, Figure <a href="#fig3">3</a> presents an illustration of the best match obtained. Moreover, the <a href="3int.html#3.1">shift-outlier</a> probabilities are presented in <a href="#tab2">Table 2</a>. Of course, the posterior distributions and graphs in Figures 1, 2 and 3 need to be correctly interpreted, together with the posterior probabilities in Table 2. The reader is referred to the <a href="3int.html">intuitive</a> section for a more general explanation and to the <a href="4tech.html">technical</a> section for greater detail.</P> <a name="fig1"></a><p class="image"><img src="images/ulandrykpostnrr.gif" alt="'Dot plot' distribution chart"><br> Figure 1: Posterior distribution for θ<sub>0</sub>, for data in Table 1, not considering the errors in the calibration curve</p> <a name="fig2"></a><p class="image"><img src="images/ulandrykposterr.gif" alt="Dot plot' distribution chart"><br> Figure 2: Posterior distribution for θ<sub>0</sub>, for data in Table 1, considering the errors in the calibration curve</p> <P>The output explained here is the only output produced by Bwigg. In section 5 of the output page the files produced by Bwigg may be downloaded. These include the figures in <a href="http://www.libpng.org/pub/png">png</a> format, which is compatible with nearly all popular word processing packages and may be directly included in your papers. Also, in the compressed (.zip) file there are lots of configuration files, the data file (.dets) and tables in ASCII format defining Figures 1, 2 (.hist files) and 3 (combining .ccur and .dets files). These files may be useful if you want to create your own figures; otherwise downloading the compressed file is not recommended. (For example, the ASCII tables may be imported into <em>Excel</em> to produce customised versions of the figures produced by Bwigg; see Figure 5.)</P> <a name="tab2"></a> <div class="centred"> <table border="1" cellpadding="5"> <tr bgcolor="#ccccff"><td>j</td><td>Lab. No.</td><td>Post. Prob</td><td>Bayes Fac.</td></tr> <tr><td>1</td><td>4084</td><td>0.0467</td><td>0.934000</td></tr> <tr><td>2</td><td>4083</td><td>0.0661</td><td>1.322000</td></tr> <tr><td>3</td><td>4082</td><td>0.0353</td><td>0.706000</td></tr> <tr><td>4</td><td>4081</td><td>0.0313</td><td>0.626000</td></tr> <tr><td>5</td><td>4080</td><td>0.0435</td><td>0.870000</td></tr> <tr><td>6</td><td>4079</td><td>0.0635</td><td>1.270000</td></tr> <tr><td>7</td><td>4078</td><td>0.0529</td><td>1.058000</td></tr> <tr><td>8</td><td>4077</td><td>0.0466</td><td>0.932000</td></tr> <tr><td>9</td><td>4076</td><td>0.1057</td><td>2.114000</td></tr> <tr><td>10</td><td>4075</td><td>0.0988</td><td>1.976000</td></tr> <tr><td>11</td><td>4074</td><td>0.0432</td><td>0.864000</td></tr> <tr><td>12</td><td>4073</td><td>0.0350</td><td>0.700000</td></tr> <tr><td>13</td><td>4072</td><td>0.0292</td><td>0.584000</td></tr> <tr><td>14</td><td>4071</td><td>0.0341</td><td>0.682000</td></tr> <tr><td>15</td><td>4070</td><td>0.0323</td><td>0.646000</td></tr> <tr><td>16</td><td>4069</td><td>0.0360</td><td>0.720000</td></tr> <tr><td>17</td><td>4068</td><td>0.0384</td><td>0.768000</td></tr> </table></div> <p class="image">Table 2: Approximations obtained for the posterior probabilities and Bayes factors for each determination being an outlier.</p> <a name="fig3"></a><p class="image"><img src="images/ulandrykmtc.gif" alt="Dot plot' distribution chart"><br> Figure 3: Data in Table 1, fixing θ<sub>0</sub> = 2191 BP; the best match according to Figure 2. Also the relevant part of the calibration curve is presented, with the estimated standard errors in the curve, to appreciate the wiggle match.</p>  <hr><p style="text-align:center"> <a href="1intro.html" onMouseOver="on('back');window.status='Go Back';return true" onMouseOut="off('back');window.status='';return true"> <img src="/icons/back.gif" name="back" ALT="[Back]" border="0"></a> <a href="3int.html" onMouseOver="on('fwd');window.status='Go Forward';return true" onMouseOut="off('fwd');window.status='';return true"> <img src="/icons/fwd.gif" name="fwd" ALT="[Forward]" border="0"></a> <a href="toc.html" onMouseOver="on('toc');window.status='Go to Article Table of Contents';return true" onMouseOut="off('toc');window.status='';return true"> <img src="/icons/toc.gif" name="toc" ALT="[Contents]" border="0"></a> <a href="/" onMouseOver="on('home');window.status='Go to Internet Archaeology Home Page';return true" onMouseOut="off('home');window.status='';return true"> <img src="/icons/ia.gif" name="home" ALT="[Journal Homepage]" border="0"></a> </p><p class="copy">© Internet Archaeology URL: <a href="http://intarch.ac.uk/journal/issue13/2/2frm.html" onMouseOver="window.status='If browsing offline, click on this link to return to the journal';return true" onMouseOut="window.status='';return true">http://intarch.ac.uk/journal/issue13/2/2frm.html</a> <br>Last updated: Thu Jan 23 2003</p>