Now click OK and then OK again and the analysis is performed, in the sample above, it yields this: To make it easier to see If the authors cluster properly, you might want to add the author’s name to each title, as I’ve done with the date.
#EXCEL MINITAB FOR MAC ARCHIVE#
Here, I’ve typed in “Word,” the label the Intelligent Archive created for the words, but which is now above the text titles after transposition. Select the “Case labels” by clicking in the box next to it and then selecting the correct column from the box on the left, or you can type the label into the “Case labels” box. Make the title informative, so you can keep track of what you did. You next click the Customize box so that you can enter a title for the graph and indicate what labels you want to use. You must also select Show dendrogram, as I have done below this produces the graph.
#EXCEL MINITAB FOR MAC TRIAL#
This only changes the color of the lines in the graph in this trial analysis with known texts, the colors tell me at a glance whether the four authors’ texts group properly. I’ve set the number of clusters to 4, the number of different authors in my set. It’s safe to use the ones shown below, where I’m analyzing the 800 MFW using Ward linkage, and Squared Euclidean distance you must also select “Standardize variables,” which compensates for the rapid decrease in word frequencies in a word list. In the Cluster Observations window that will pop up, first select the words you want to analyze, either by selecting them with your mouse in the box at the left or by typing the column numbers into the “Variables or distance matrix:” box (here, I’ve typed in “C2-C801” to analyze the 800 most frequent words the labels are in column 1). To do a cluster analysis of the data above in Minitab, select the Stat menu, then Multivariate, then Cluster Observations. (If you need to create a blank worksheet in Minitab, go to the File menu, select New, then Minitab Worksheet.) Your data should now look like that below. Go to Minitab and paste the data into a blank worksheet by clicking the unmarked grey cell below C1 and above row 1 and pasting the data with Ctrl-V. (You can transpose more than 1000 words, but Minitab can only analyze about 1000 at a time in cluster or PCA t-testing may involve larger numbers of words.) Once your data is transposed, select your transposed data by clicking the top left cell, going to the bottom right of the data and clicking the bottom right cell and using Ctrl-C to copy the data. Your data will now be transposed so that the authors and titles are on the left and the words are along the top. When the Paste Special dialogue box comes up, select Transpose (just above the OK button) and then click OK. Next, copy the data with Ctrl-C, then click on an empty cell to the right of your data, select Paste from the Home menu, then Past Special. Now select the data in the sample above, click cell A1, then scroll down to the last row, press and hold the Shift key and click the bottom right cell. Click the grey cell above row 1 (with just “A” in it) to select the entire column 1, then use Ctrl-F (or Find & Select then Replace, from the Home menu) to find all apostrophes and replace them (I use “`”). To use the words in your analysis as column labels, first replace any apostrophes with another character Minitab will not allow apostrophes in labels. If you are using Excel 2007 or later, you can prepare the data in Excel. Data Preparation: before you move the data into Minitab for analysis, you need to prepare it a bit. Once you have created your word lists, you should have data that looks like that below.
If you use the Intelligent Archive, be sure to change frequencies to proportions before copying the data. A more elegant and much faster method is to use Hugh Craig’s Intelligent Archive (available here with instructions). The spreadsheet also lets you delete personal pronouns, and/or remove words found in the word list only because they are extremely frequent in one text (see the instructions, and below, for more discussion). One method for creating these word lists is to use the Delta Spreadsheet (available, with detailed instructions on this site). That is, that there is a single master list of words, and that each text has an entry for each word, whether or not the word is found in that text. Note: All of these instructions assume that you have, or can create, a set of parallel word lists for a group of texts.
Cluster Analysis, Principal Components Analysis (PCA), and T-testing in Minitab ©, and Excel ©
0 kommentar(er)
