BIOLOGY 458 LAB 1 - AN INTRODUCTION TO SPSS AND R INTRODUCTION

Transcription

1 BIOLOGY 458 LAB 1 - AN INTRODUCTION TO P AND R INTRODUCTION This course was originally designed to use the statistical software P as a computational aid. With the advent of the comprehensive open source package R, I have attempted to provide additional instruction to permit those inclined to use R. The sole advantage to R is that it is free and available on all platforms (Windows, MAC, and Linux). If you learn to use R, you can easily port that knowledge to other Universities or into business or industry, whereas you might find access to P limited when you move elsewhere. The disadvantage to R is that it is a largely command-line oriented package written by experts, predominantly for experts. To obtain R, go to You will need to first choose your CRAN mirror, then you can download the binaries of R for the operating system you plan to use. Follow the instructions to install R on your computer. Besides the base level R package, there are many user contributed additional modules for R. As the labs progress I will advise you on additional modules to download and install, and how to do so. As for a manual on R, my advice is to use Google to find advice on R. The internal R documentation is marginally helpful. o what follows here for each lab exercise will be instruction for use of P, and then at the end of each lab section additional advice on how to obtain the necessary functionality in R to complete the lab exercise. I will try to demonstrate both P and R during the class period, and additionally during the in class lab period P (statistical package for the social sciences) is one of many statistical software packages available today. It is widely used by scientists and social scientists since it is a comprehensive package of numerous procedures for data analysis and interpretation. Mainframe, PC, and MAC versions of P are available. Throughout this course, we will be learning how to use the PC or MAC versions of this program. P is a menu driven "point and click" software package. When you first enter P, the screen you will see is called the Data Editor Window.

2 A "toolbar" with pull-down menus for various commands is found at the top of the Data Editor Window. On the right-hand end of the toolbar in the Data Editor Window is the Help Menu. Like most other software packages P offers extensive explanatory information. In addition to the encyclopedic list of help topics which may be accessed from the Topics command in the Help Menu, P also offers a Tutorial, and a tatistics Coach which are both accessed from the Help Menu. The Tutorial is an interactive introduction to numerous aspects of using P. It may be entered at a number of different points, so it is useful anytime you are attempting to use P in a novel way. The tatistics Coach is also an interactive guide to data analysis and interpretation. P asks you a series of questions and your answers lead the P tatistics Coach to recommend what procedures to apply to answer the question you pose for the data you have. Point To Remember: While P offers an amazing array of options for data transformation, analysis, graphing, and output, a only a small number of commands need to be mastered to find P useful immediately. I personally do not know all the bells and whistles of P, particularly the Windows based interface since I have used

3 mainframe and DO versions of P for many years. I learn more as I need to use P in new ways. Feel free to explore this program as much as you like! Editorial Note: I have attempted to put the names of the various Menus in P in bold and italic and the specific Commands available in a menu in bold and underline. GETTING TARTED I have found the File, Transform, tatistics, Graphs and Help Menus to be those that I use most frequently in P. The five main steps in an P session are Enter the Data Transforming Data Values Analyzing Data Graphing Data Examining, Printing, and aving Results Opening or Reading Data Files Data Entry Normally, the first item of business in any P session is to open a data file, read in a data file, or to type in your data directly into the Table in the Data Editor Window. Data files may be "opened" using the OPEN command if they are P files with the extension (.sav) or using the OPEN DATABAE command if they are spreadsheet files. They are read in using the READ TEXT DATA command if they are ACII, Text or DO files. The commands Open, OPEN DATABAE, and READ TEXT DATA are found in the File Menu. I have had less success opening Excel spreadsheet files using the instructions provided in the Help Menu, than I had transferring data into P from Excel spreadsheets by highlighting the data I wanted to transfer, copying it to the clipboard, and pasting it into P. The data can then be saved as an P file for future analysis.

4 Typing Data into the P Data Editor Window You can enter your data directly into the Table in the P Data Editor Window. Each line in the Table represents a different experimental subject and each column a different variable. You can use integer codes to represent particular treatment levels. Variable names can be entered into the Table by double-clicking the left mouse button on the variable name box at the top of a column and typing in a name in the appropriate subwindow. Defining Missing Values Often we are unable to collect some data on particular experimental subjects. When this occurs, we usually enter some value into the data file to hold the place of this "missing observation." The value could be a set of blank spaces or a numeric value that is easy to distinguish from the actual observations. For example, one might enter the value "-1" for data on species diversity. ince species diversity must always be a positive real number, missing observations assigned the value "-1" are easy to distinguish from actual observations. In P in the DATA EDITOR WINDOW, click on the command VARIABLE VIEW tab at the bottom of the window to see the options for defining the attributes of each of your variables. Note that one option is to define the value you use to indicate a missing value.

5 electing ubsets of the Data for Analysis ometimes we wish to only select a subset of the experimental subjects or to select a subset of the data from the entire range of values a particular variable may cover. For example, if we have data in which one column of the data file contains integer codes for sex with 1 indicating male and 2 indicating female, we may wish to perform an analysis or graph the data for females only. In the Data Menu, click on the command elect Cases. Transforming Data Values P offers a number of options for transforming data values. However, I have found that the Compute and Recode commands under the Transform Menu are the ones I use most often.

6 Compute - The Compute command allows you to perform a wide variety of numerical transformations to your data. imply click on the Compute command, enter the name of the new variable you wish to create in the Target Variable sub-window, click on the name of an existing variable and the right arrow button to move that variable to the Numeric Expression sub-window, and then use the calculator like keypad or the Functions sub-window and button to construct your data transformation as an equation. Click on OK to have the transformation performed. P will transform the data according to the equation you entered, and place the new variable in the next empty column in the data file or Table. Recode - The Recode command is useful as a means of redefining data values for the purpose of aggregating data from different categories or for replacing data values coded in ways that P does not like. For example, if we had an additional variable in our data file that categorized animals into age categories, with 1 indicating infants, 2 indicating juveniles, 3 indicating sub-adults, 4 indicating sexually active adults, and 5 indicating non-sexually active adults, we might wish to perform an analysis that lumps all adult and sub-adult categories into just adults and non-adults. The Recode command would allow us to redefine values of the variable age to achieve this result. imply click on the Transform Menu and the command Recode. Note that one can recode the existing data, that is, place the recoded values in the age column of the data file in place of the old codes, or create a new variable in the next empty column of the data file or Table, with the recoded values as a newly created variable. I tend to do the latter, since I often screw-up the recoding and need to repeat the process several times before I achieve my desired recoding. Again the exact process is a mouse and button clicking helluva good time. Analyzing Data In P 10.0 the Analyze Menu is where one finds all the procedures in P for the analysis of data. The commands Descriptives and Custom Tables are used to

7 compute various summary statistics, such as means, variances, standard deviations, standard errors, skewness, kurtosis, and others for particular variables for all your data or for subsets of your data. Under Descriptives, the option Explore is particularly useful. The arrows on the right side of the Analyze menu indicate that more options are available by following the arrow with the mouse and cursor. The commands Compare Means, General Linear Model, and Nonparametric Tests are used for most analyses in which the responses of subjects to a variety of experimental treatments or other categorical conditions are contrasted. The commands Correlation and Regression are used when examining the relationships between two or more continuous variables observed on each subject. These are the primary commands concerned with data analysis that we will use during this course. Graphing Data P provides a variety of ways to graph your data for visual presentations mostly from the Graphs Menu, and considerable ability to edit and customize your graphs to suit your specific needs. Many of the procedures accessed from the Analyze Menu also have the option of generating graphs. You can save graphs to a disk file or print them out.

8 Examining, Printing, and aving Results The results of the application of any of the statistical procedures or graphing procedures from the Analyze or Graphs Menus appear in the P Output Navigator Window. The Output Navigator Window is divided into two parts. The left side of this window is an Output Tree which guides you through the various parts of your results. It may include a log of all P commands that were executed in the P Log. The results of each procedure executed and subsets of those results are also shown in the output tree as seen in the example below for the command Graph - catter. One may access various parts of the output by double clicking on the icons in this tree, and one may suppress display of parts of the output by double clicking on an icon. An open book icon means the output is displayed and a closed book means it is not. A scroll bar is provided to allow you to move up and down the Output Tree. The right side of the Output Navigator Window shows you the actual results of the data analysis or graphing procedure you have executed. You may use the scroll bar at the right side of this window to move up and down in the output. In addition, you can edit items in the output window such as captions on graphs and axis labels, or enter additional text or notes. These kinds of procedures are accomplished by clicking directly on the objects in the output window, by using commands in the Edit Menu, or by clicking on particular icons in the toolbar.

9 After generating output you may save it to a disk file or send it to a printer. These tasks are accomplished from the Output Navigator Window using commands on the File Menu. A Point to Remember: All of this will become clearer as you use trial and error to get accustomed to P and its user interface. GETTING DOWN TO BUINE As a brief introduction to using P, complete the following exercise during the lab period. But, feel free to play with P to get accustomed to its capabilities and its user interface. Feel free to use you own data to continue your exploration of P, either during the lab period or in the public access labs at other times. LAB EXERCIE The following lab exercise is intended to allow you to become familiar with the P Menu/Help system and to create analyze some real data. You will learn how the commands function using the on-line Menu/Help system. A few words about the data set:

10 The data for this week's lab come straight from some male-male competition work that a former student collected for a species of crab spider, Miseumna formosipes (Family: Thomisidae). It is in the file malrank, of which you will need to obtain a copy. The file contains data on specific spider males including: their code number (Column 1), carapace width in mm (Column 2), tibia length mm (Column 3), body length mm (Column 4), and weight in mg (Column 5). Another important feature of the data file is that weights for some the spiders were not recorded. In order for P to read the data format, a value is required. In these cases I often use to indicate missing values. Keep this in mind for any analyses that use the variable "weight." Again, the first thing to do in a program is to tell P where to find the data, what form it is in, and names for the variables. In this case you have several options. You can have P Read Text Data, but you will need to indicate the file name, its location, and how the data are delimited. Next, obtain the mean and standard deviation for all the variables (except spider code obviously). There are several ways to do this, but the easiest commands are accessed through the Analyze Menu - Descriptive tatistics (Descriptives or Explore) and Custom Tables. You will need to define the missing values for the variable "weight", otherwise P will treat the values of as valid entries. You will need to exclude these values by using the Variable View tab in the Data Editor Window and defining the missing value to be Now make a plot (a scatter plot) of weight versus carapace width. Make sure to make weight the dependent variable (y-axis). This is found within the menu system under Graph Menu. Be sure to label the axes (vertical and horizontal), give the graph a title, and to remind P that you have some missing data. Next make a histogram of tibia lengths and superimpose a normal distribution curve over the histogram. Again look to the Analyze Menu for the Explore command, or the Graph Menu for the Histogram command. Further Instructions for the Lab Exercise The data file for lab 1 malrank is linked to the webpage. imply click on the link and then save the file on your computer or mass storage unit (disk or UB drive). This makes the file local to your computer and you can then use the Read Text Data command and Wizard in P to read the data into P. After completing the exercise, you can save your results by using the elect All command from the Edit Menu, and then the Copy Objects command from the Edit menu to copy your results to the clipboard. In the mean time, open a Word file and then Paste the results into the Word file. Name and save your Word file, and copy it to your mass storage unit, or it to yourself (or both). Now you can work on your lab writeup at home even if you don t have access to P. If you have P on your own computer, you can simply use the ave As command from the File Menu to save both

11 your results from the Output Viewer window and the P version of the data file from the Data Editor Window. Make sure that your computer at home or the office has at least as recent a version of P as we use in lab, or the files may be incompatible. It is possible to save P files in a format compatible with earlier versions of P. Lab Write-Ups Lab write-ups should be submitted as Word or RTF files. Name the file yourlastnamelab1 for example. Lab write-ups should include answers to any questions posed by the lab exercise and report and interpret the results of any analysis performed. Lab write-ups that consist only of P output without a narrative providing answer, results and interpretation will be treated as if nothing was submitted. Not all of the tables and graphs you produce in a work session need to be turned in for any particular lab write-up. Only the tables and graphs you deem essential to convince me that you have mastered using P to complete the exercise should be turned in. No lab write-up need be turned in for lab 1. However, I have included below an example of what a write-up for lab 1 might look like. Take this as a lesson for what is expected of you for labs 2-10, all of which require a written lab write-up to be submitted. Edward F. Connor Lab 1 Write-Up Using the data provided on body size and mass for male crab spiders (Miseumna formosipes), I calculated the means and standard deviations of each metric variable using the Descriptives procedure in P (see table below) Of the linear measurements body length was most variable (sd = 0.307), but also had the highest mean value. Total body mass (weight) averaged 6.4 mgs, but was also quite variable among individual spiders. Only 23 of the 38 spiders had data on body mass.

12 Descriptive tatistics carawid tibalen bodylen weight Valid N (listwise) N Mean td. Deviation I then proceeded to produce a scatter plot of body weight versus carapace width (see graph below). The graph suggestions that weight increases linearly with body mass for these male spiders A histogram of tibia lengths suggestions that tibia length in mm is approximately normally distributed, although there is some evidence that the distribution is slightly skewed to the right (the distribution is not symmetrical).

13 LAB 1 - AN INTRODUCTION TO R Note usable lines of syntax are labeled and indented Text in italic is actual output from R Installing Accessory Packages in R To install and load accessory packages in R, you must be connected to the internet. Open R on your computer. Click on the Packages menu at the top of the R-Gui. Click on Install Packages. You may be asked to select your CRAN site. Choose CA1. Next, highlight the package you want and click OK at the bottom of the window that lists all the available packages. Once the package is installed on your computer, you will still need to load the package each time you want to use it. Load packages with this command: library(package name) Reading in Data Files in R

14 Using the sample data file pit.csv (pit.csv) which has in the first column integer codes for treatment (treat) and in the second column the data to be analyzed (funct) To read in a tab delimited text data file with variable names use: Create an object name such as data1 data1<-read.csv( c:\\pit.csv, header=t) Alternately, if you do not know the file path (or don t feel like typing it in) you can use this command. It will allow you to browse for the data file you want. data1<-read.csv(file.choose(), header=t) To read an P data file into R use this command you must use the foreign package library(foreign) Then you should be able to read in the file. You do not need to specify that there are headers. data1<-read.spss(file.choose()) Type data1 to see the data file data1 To make each variable (column) accessible individually use: attach(data1) Alternatively concatenate the two commands with a semicolon separator as: Or data1<-read.csv( c:\\pit.csv, header=t);attach(data1) data1<-read.csv(file.choose(), header=t);attach(data1) Useful R Package-Rcmdr (R-Commander) R-Commander is a Graphical User Interface (GUI) for R with pull-down menus for data manipulations, statistical tests, and other R functionality. It is especially useful for those who have never used R before and are unfamiliar with the syntax used in the R console. When using the pull down menus R-commander will show you the R console syntax for the command you just used, so it can also help you learn R console syntax as you use it. You can download R commander from the CRAN Website.

15 For an excellent R-commander tutorial, go to the following website: Making Data from R-Commander Available in the R-Gui Once you have read a datafile into R- commander, you can examine the columns in that data file as individual variables from the R-Gui by executing the command attach(datafile) (substitute the name of the active dataset for datefile) attach(datafile) Type datafile to see the datafile datafile Now that you have attached the datafile, the column names are accessible as variable names for analysis. Useful R Package-ggplot2 GGplot 2 is a package for creating graphics with R. The syntax is fairly straightforward and you can generate professional looking graphics on par with programs like igma Plot, Excel, and KaleidaGraph. GGplot2 can be downloaded at the CRAN website. For a good ggplot2 tutorial, visit this website: Examples of GGplot2 plotting code: Loading a package that you have already downloaded library(ggplot2) Loading data into R from csv file size<-read.csv(file.choose()) Making simple boxplot using ggplot2 (aes = short for aesthetics) p=ggplot(size, aes(x=factor(site), y = L)) p+geom_boxplot() Adding complexity

16 assign the fill colour of the boxes to represent different cruises p=ggplot(size, aes(x=factor(site), y = L)) p+geom_boxplot(aes(fill=factor(cruise)) Label the colour legend and label each colour with a date p+geom_boxplot(aes(fill=factor(cruise))+ scale_fill_discrete( Cruise, breaks=c(5,6,7,8,9,10), labels=c( Mar, Apr, May, June, July, ept )) change y and x axis labels and change scale limits on y axis p+geom_boxplot(aes(fill=factor(cruise))+ scale_fill_discrete( Cruise, breaks=c(5,6,7,8,9,10), labels=c( Mar, Apr, May, June, July, ept ))+scale y_continuous( Clam Length (mm), limits=c(5,22))+xlab( ampling Location ) EXTRA! code for adding colours by number within scale_colour or scale_fill values=c(colours()[c(153, )]) Making a simple scatterplot p=ggplot(classdata1,aes(x = L, y = W)) p+geom_point() Adding complexity p=ggplot(classdata1,aes(x = L, y = W)) p+geom_point(aes(colour=factor(ite))) add a smoothed line with 95% CI p+geom_point(aes(colour=factor(ite)))+geom_smooth() change color of the line and making it a linear regression m ) p+geom_point(aes(colour=factor(ite)))+geom_smooth(colour= black,method= l remove the 95% CI shading and change line type

17 p+geom_point(aes(colour=factor(ite)))+geom_smooth(colour= black, method= lm, se=f, linetype=5) add separate lines (linear regressions) for each site and make the line colour match the colour of the points p+geom_point(aes(colour=factor(ite)))+geom_smooth(aes(group=factor(ite), colour=factor(ite)), method= lm, se=f ) Plot with means with errorbars Basic p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_point() Adding complexity assigncolour of points to represent different sites p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_point(aes(colour=factor(ite))) same plot, change the size of the points p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_point(aes(colour=factor(ite)),size=4) add error bars using standard error from classdata2 p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_point(aes(colour=factor(ite)),size=4)+geom_errorbar(aes(ymax=mean L+seL, ymin=meanl-sel)) put error bars behind the points and change their width p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_errorbar(aes(ymax=meanl+sel, ymin=meanlsel),width=0.05)+ geom_point(aes(colour=factor(ite)),size=4) Change colour of errorbars p=ggplot(classdata2, aes(x=factor(cruise), y = meanl))

18 p+geom_errorbar(aes(ymax=meanl+sel, ymin=meanlsel),width=0.05, colour= dark grey)+ geom_point(aes(colour=factor(ite)),size=4) make the colour of the errorbars match the point colour p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_errorbar(aes(ymax=meanl+sel, ymin=meanlsel, colour=factor(ite)),width=0.05)+ geom_point(aes(colour=factor(ite)),size=4) Adding x and y labels, and labels to color legend p=ggplot(classdata2, aes(x=factor(cruise), y = meanl)) p+geom_errorbar(aes(ymax=meanl+sel, ymin=meanlsel, colour=factor(ite)),width=0.05)+ geom_point(aes(colour=factor(ite)),size=4)+scale_colour_discrete( ite, breaks=c(1,2,3,4), labels=c( ite 1, ite 2, ite 3, ite 4 ))+scale_y_continuous( Mean Clam Length (mm), limits=c(6,16))+ scale_x_discrete( ampling Date, breaks=c(5,6,7,8,9,10), labels=c( Mar, Apr, May, June, July, ept ))