The easiest way to use provenance without the need to know the syntax of the R programming language is to start R and type the following commands at the prompt: 
library(provenance)
provenance()
This brings up the following menu, from which the functions of interest can be chosen:

Tutorials:
Example 1: QFL diagrams
Example 2: Radial plots
Example 3: Kernel Density Estimates (KDEs)
Example 4: Multi-sample, multi-method summary plots
Example 5: Source Rock Density (SRD) correction
Example 6: Hydraulic sorting of heavy minerals
Example 7: Multidimensional Scaling (MDS)
Example 8: Principal Component Analysis (PCA)
Example 9: Correspondence Analysis (CA)
Example 10: 3-way MDS

Example 1: Plotting a QFL diagram

Choose option 2 from the start menu: 
Pick an option:
1 - sample size calculation
2 - plot a single dataset
3 - plot multiple datasets
4 - Minsorting
5 - MDS/PCA/CA
6 - Procrustes analysis
7 - 3-way MDS
8 - save plots (.pdf)
9 - help
q - quit
2
This brings up a new list of options. Select the first of these: 
Plot a single dataset:
1 - Ternary diagram
2 - Pie charts
3 - Radial plot
4 - Cumulative Age Distributions
5 - Kernel Density Estimates
1
Next you are asked to choose between two data types: 
1 - Load a compositional dataset
2 - Load a point-counting dataset
The petrographic data are reported as raw counts and have not been normalised to a common sum. So the second option is the one to choose here. If your counting data are reported as fractions or percentages, then you must choose the first option. 
Open a compositional dataset:
Enter file name: PT.csv
This loads a petrographic dataset containing six different classes: quartz (Q), K-feldspar (KF), plagioclase (P), and lithic fragments of metamorphic (Lm), igneous/volcanic (Lv) and sedimentary (Ls) origin. Next, we will amalgamate these six classes into three categories by selecting the third option in the following list: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
3
This brings up a nested sequence of queries, in which three new amalgamated groups are defined: quartz (Q), feldspar (KF + P) and lithics (Lm + Lv + Ls). The amalgamation is concluded by entering an empty line: 
Select a group of components from the following list:
Q,KF,P,Lm,Lv,Ls
Enter as a comma separated list of labels or click [Return] to exit:
Q
Name of the amalgamated component? quartz
Select a group of components from the following list:
quartz,KF,P,Lm,Lv,Ls
Enter as a comma separated list of labels or click [Return] to exit:
KF,P
Name of the amalgamated component? feldspar
Select a group of components from the following list:
feldspar,quartz,Lm,Lv,Ls
Enter as a comma separated list of labels or click [Return] to exit:
Lm,Lv,Ls
Name of the amalgamated component? lithics
Select a group of components from the following list:
lithics,feldspar,quartz
Enter as a comma separated list of labels or click [Return] to exit:
Three other functions can be combined with the amalgamation operation, but for the sake of this tutorial, we will not use them and proceed by entering 'c': 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
Since our amalgamated dataset contains quartz, feldspar and lithic components, it is useful to plot it on an QFL diagram: 
Plot background lines?
1 - basic grid [default]
2 - descriptive QFL diagram
3 - Folk's classification
4 - Dickinson's QFL diagram
5 - no lines
2
Finally, let's add a confidence ellipse around the data. Accept the default value to obtain a 95% confidence region: 
1 - Add an error ellipse the entire population
2 - Add an error ellipse for the average composition
3 - No ellipse [default]
1
Confidence level [default=0.05]?
This calculation may take up to a minute, because it involves the repeated numerical solution of a double integral. Here is the output:

Example 2: Radial plot

Point-counts are affected by a combination of (1) true compositional variability and (2) multinomial counting uncertainties. The relative significance of these two sources of dispersion can be visually assessed on a radial plot. The following tutorial will use this graphical device to assess the dispersion of the garnet-to-epidote ratio in Namib desert sand. Select the second option from the menu: 
Pick an option:
1 - sample size calculation
2 - plot a single dataset
3 - plot multiple datasets
4 - Minsorting
5 - MDS/PCA/CA
6 - Procrustes analysis
7 - 3-way MDS
8 - save plots (.pdf)
9 - help
q - quit
2
Choose the third option and open the heavy mineral dataset: 
Plot a single dataset:
1 - Ternary diagram
2 - Pie charts
3 - Radial plot
4 - Cumulative Age Distributions
5 - Kernel Density Estimates
3
Open a point-counting dataset:
Enter file name: HM.csv
Accept the default options for the next step: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
Next we are presented with all the components of HM.csv We must choose two of these components to form a ratio. Let's select epidote (ep) and garnet (gt): 
Select two components from the following list to 
form the numerator and denominator of the ratios to be displayed:
zr,tm,rt,TiOx,sph,ap,ep,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated pair of labels:
ep,gt
The resulting radial plot contains a lot of useful information. First, the data scatter significantly beyond a 2-sigma error band around the origin, indicating that the data are overdispersed with respect to the point-counting uncertainties. This is also reflected in the MSWD-value, which is significantly greater than 1. Equivalently, the data fail the chi-square test for compositional homogeneity, with a p-value of close to zero. The excess scatter of the data can be quantified using a random effects model with 74% dispersion. This estimates the true geological variability of the ep/gt-ratios, with the counting uncertainties removed. Thus, the true ep/gt-ratios are described by a lognormal distribution with a geometric mean of 0.887 ± 0.046 (the 'central ratio') and a coefficient of variation (standard deviation divided by the mean) of 0.74.

Example 3: Plot a Kernel Density Estimate (KDE)

In this tutorial we will plot a single detrital zircon U-Pb age distribution as a KDE. After selecting the second option from the main menu, we pick the fifth choice and open the DZ.csv datafile: 
Plot a single dataset:
1 - Ternary diagram
2 - Pie charts
3 - Radial plot
4 - Cumulative Age Distributions
5 - Kernel Density Estimates
5
Open a distributional dataset:
Enter file name: DZ.csv
For the sake of this exercise, we will just plot a single sample (N13), so we need to subset the multi-sample dataset. Instructions for plotting multiple samples are given in Example 3: 
Options:
1 - Subset samples
2 - Load analytical uncertainties
c - Continue
1
Select a subset group of samples from the following list:
N1,N2,N3,N4,N5,N6,N7,N8,N9,N10,N11,N12,N13,N14,T8,T13
Enter as a comma separated list of labels:
N13
We accept the default settings for all remaining options: 
Options:
1 - Subset samples
2 - Load analytical uncertainties
c - Continue
c
Options:
1 - Set minimum age
2 - Set maximum age
3 - Turn off adaptive density estimation
4 - Plot on a log scale
5 - Set bandwidth
c - Continue
c
Which produces the following graphical output:

Example 4: Multi-sample, multi-method summary plot

In this tutorial, we will plot a large 15-sample, 5-method dataset from Namibia including: First we need to load these samples. Let's start with the major element compositions: 
1 - Add a compositional dataset
2 - Add a point-counting dataset
3 - Add a distributional dataset
c - Continue
1
Open a compositional dataset:
Enter file name: Major.csv
We get presented with a new selection menu. Let's accept the default settings, which brings us back to the previous menu: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
1 - Add a compositional dataset
2 - Add a point-counting dataset
3 - Add a distributional dataset
c - Continue
Repeat the same steps to load Trace.csv. Next, we will load the two point-counting datasets: 
1 - Add a compositional dataset
2 - Add a point-counting dataset
3 - Add a distributional dataset
c - Continue
2
Open a point-counting dataset:
Enter file name: PT.csv
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
Repeat for the trace element element compositions (Trace.csv). Finally, we load the detrital zircon age distributions: 
1 - Add a compositional dataset
2 - Add a point-counting dataset
3 - Add a distributional dataset
c - Continue
3
Open a distributional dataset:
Enter file name: DZ.csv
Distributional data are plotted as Kernel Density Estimates, which can be modified by any of seven options. For this tutorial, we will again accept the default settings: 
Options:
1 - Set minimum age
2 - Set maximum age
3 - Turn off adaptive density estimation
4 - Plot on a log scale
5 - Set bandwidth
6 - Use the same bandwidth for all samples
7 - Normalise area under the KDEs
c - Continue
c
Type 'c' again to exit the file selection menu and generate a two-column summary plot: 
1 - Add a compositional dataset
2 - Add a distributional dataset
c - Continue
c
Number of columns? 2
Which produces the following output:

Example 5: Source Rock Density (SRD) correction

The effect of hydraulic sorting on the petrographic and heavy mineral composition of detrital suites can be undone by restoring their density to an assumed value. This procedure works because, a few exceptions notwithstanding, the average density of continental crust varies very little (between 2.65 and 2.8 g/cm3, say) over all but the smallest catchment areas. In this tutorial, we will visualise the SRD correction on a ternary diagram. Like in the first tutorial, we select the second option in the main provenance() menu, and open a file (PTHM.csv) containing the relative proportions of all the detrital components (light plus heavy minerals): 
Plot a single dataset:
1 - Ternary diagram
2 - Pie charts
3 - Cumulative Age Distributions
4 - Kernel Density Estimates
1
1 - Load a compositional dataset
2 - Load a point-counting dataset
1
Open a compositional dataset:
Enter file name: PTHM.csv
Note that we chose the first option because the data are expressed as percentages and not counts. In order to apply the SRD correction, we need to provide a table of mineral densities. provenance comes preloaded with a set of default densities. We can either use these or load a different table. It is important to make sure that the density table uses the same category labels as the file with the sample compositions. Finally, we need to provide the assumed density of the source rock. For this tutorial, we will assume a 2.71 g/cm3: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
1
Open a density file [y] or use default values [N]? n
Enter target density in g/cm3: 2.71
Since we want to plot the SRD corrected composition on a ternary diagram, we need to reduce the number of components in the dataset from 26 to 3. This can be done by amalgamation (see the first example) or subsetting, as shown next. Let's select garnet (gt), epidote (ep) and amphibole (amp): 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
2
Select a subset group of components from the following list:
Q,KF,P,Lv,Ls,Lm,mica,opaques,FeOx,turbids,zr,tm,rt,TiOx,sph,
ap,ep,othLgM,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated list of labels:
ep,gt,amp
Finally, we plot the SRD-corrected gt-ep-amp composition on an empty ternary diagram: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
Plot background lines?
1 - basic grid [default]
2 - descriptive QFL diagram
3 - Folk's classification
4 - Dickinson's QFL diagram
5 - no lines
1
Show SRD correction [Y or n]? y
Where the last instruction adds lines to the diagram, showing the effect of the SRD correction on the ep-gt-amp composition. Finally, let's skip the error ellipse calculation: 
Show SRD correction [Y or n]? 
1 - Add an error ellipse the entire population
2 - Add an error ellipse for the average composition
3 - No ellipse [default]

Example 6: Hydraulic sorting of heavy minerals

In this tutorial, we will compute the grain size distribution of heavy minerals in a sediment sample. To this end, select option 4 from the main provenance() menu. Choose the first option of the next menu to model the heavy mineral composition of an actual sample: 
Change default:
1 - bulk composition [default=tectonic endmembers]
2 - average grain size in phi units [default=2]
3 - grain size standard deviation [defaults=1]
4 - transport medium [default=seawater]
5 - plot resolution [default: from -2.25 to 5.5 by 0.05]
6 - selected minerals [default: all]
c - continue
1
1 - Load a compositional dataset
2 - Load a point-counting dataset
2
Open a point-counting dataset:
Enter file name: HM.csv
Rather than plotting all 15 minerals, select zircon (zr) and amphibole (amp): 
Change default:
1 - bulk composition [default=tectonic endmembers]
2 - average grain size in phi units [default=2]
3 - grain size standard deviation [defaults=1]
4 - transport medium [default=seawater]
5 - plot resolution [default: from -2.25 to 5.5 by 0.05]
6 - selected minerals [default: all]
c - continue
6
Select a subset group of components from the following list:
zr,tm,rt,TiOx,sph,ap,ep,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated list of labels:
zr,amp
Type 'c' and choose sample N1: 
Change default:
1 - bulk composition [default=tectonic endmembers]
2 - average grain size in phi units [default=2]
3 - grain size standard deviation [defaults=1]
4 - transport medium [default=seawater]
5 - plot resolution [default: from -2.25 to 5.5 by 0.05]
6 - selected minerals [default: all]
c - continue
c
Select one of the following samples to plot:
N14,N13,N12,N11,N10,N9,N8,N7,N6,N5,N4,N3,N2,N1,T8,T13
or press [Return] to exit
N1
Which produces the following plot, showing that fine grained zircon is hydraulically equivalent to coarser grained amphibole in the sample:

Example 7: Multidimensional Scaling (MDS)

In this tutorial we will perform an MDS analysis of the detrital zircon dataset. After selecting the fifth option from the main menu, we open the DZ.csv file: 
1 - Load a compositional dataset
2 - Load a point-counting dataset
3 - Load a distributional dataset
3
Open a distributional dataset:
Enter file name: DZ.csv
We will use the default Kolmogorov-Smirnov dissimilarity for this exercise. Alternatively, it is also possible to use the Sircombe-Hazelton distance but this requires a separate input file with analytical uncertainties. This can be done by selecting the third option in the following menu. For this exercise, however, we will accept the default settings: 
Options:
1 - Subset samples
2 - Combine samples
3 - Load analytical uncertainties
c - Continue
c
Choose:
1 - Kolmogorov-Smirnov distance [default]
2 - Kuiper distance

initial  value 10.608343 
iter   5 value 8.868203
iter   5 value 8.867183
iter  10 value 7.883745
iter  15 value 7.637849
final  value 7.619237 
converged
provenance() has calculated a preliminary MDS-configuration, using a non-metric MDS algorithm. If the stress value is extremely low, the user is strongly encouraged to switch to classical scaling in order to avoid overfitting the data. The stress is 7.64% in our current example, and so the classical option is not presented here. Finally, we can modify the visual appearance of the MDS configuration in several ways. In this example, we will add solid lines connecting nearest neighbours in Kolmogorov-Smirnov space (and dashed lines connecting second-nearest neighbours), increase the size of the plot symbols, and accept the default settings for all other options: 
Options:
1 - Add nearest neighbour lines
2 - Change plot character
3 - Change size of plot character
4 - Change position of text label relative to plot character
5 - Add X and Y axis ticks
c - Continue
1
Options:
1 - Add nearest neighbour lines
2 - Change plot character
3 - Change size of plot character
4 - Change position of text label relative to plot character
c - Continue
3
Magnification of the default plot character [1 = normal]: 4
Options:
1 - Add nearest neighbour lines
2 - Change plot character
3 - Change size of plot character
4 - Change position of text label relative to plot character
5 - Add X and Y axis ticks
c - Continue
c
This produces two pieces of graphical output: an MDS configuration and a 'Shepard plot' illustrating the goodness of fit of the non-metric configuration:

Example 8: Principal Component Analysis (PCA)

The previous tutorial showed how to construct an MDS configuration for a distributional dataset. Nearly exactly the same procedure can be applied to compositional datasets such as heavy mineral compositions or chemical compositions. In strictly positive compositional datasets, for which the Aitchison distance can be safely used, it is also possible to perform Principal Component Analysis (PCA), which is a special case of classical MDS. PCA offers the advantage that it can jointly show the configuration and the endmember compositions as a 'biplot'. The current tutorial will illustrate this using the major element composition of the Namib dataset. First, we select the fifth option of the main provenance() menu. 
Pick an option:
1 - sample size calculation
2 - plot a single dataset
3 - plot multiple datasets
4 - Minsorting
5 - MDS/PCA/CA
6 - Procrustes analysis
7 - 3-way MDS
8 - save plots (.pdf)
9 - help
q - quit
5
Next we load the Major.csv file: 
1 - Load a compositional dataset
2 - Load a point-counting dataset
3 - Load a distributional dataset
1
Open a compositional dataset:
Enter file name: Major.csv
Then we accept the default settings: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
At this point, we are offered three options, the first of which performs PCA analysis: 
1 - Use Aitchison distance (PCA, default)
2 - Use Aitchison distance (MDS)
3 - Use Bray-Curtis dissimilarity (MDS)
1
The resulting graphical output is shown below:

Example 9: Correspondence Analysis (CA)

To carry out a CA of point-counting data is very similar to doing a PCA of compositional data. We begin by choosing the fifth option of the main menu: 
Pick an option:
1 - sample size calculation
2 - plot a single dataset
3 - plot multiple datasets
4 - Minsorting
5 - MDS/PCA/CA
6 - Procrustes analysis
7 - 3-way MDS
8 - save plots (.pdf)
9 - help
q - quit
5
Load the heavy mineral compositions file as a point-counting dataset: 
1 - Load a compositional dataset
2 - Load a point-counting dataset
3 - Load a distributional dataset
2
Open a point-counting dataset:
Enter file name: HM.csv
Our dataset contains 15 variables (minerals), which is a lot compared to the 16 samples. Most of these variables are dominated by zeros. Although CA is specifically designed to handle zeros, it is best reduce their impact by amalgamating some components and selecting some others. Let us begin by amalgamating the ultra-stable minerals zircon, rutile and tourmaline. 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
3
Select a group of components from the following list:
zr,tm,rt,TiOx,sph,ap,ep,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated list of labels or click [Return] to exit:
zr,tm,rt
Name of the amalgamated component? ztr
We don't want to amalgamate any further components, so let's hit [Return] to exit the amalgamation function: 
Select a group of components from the following list:
ztr,TiOx,sph,ap,ep,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated list of labels or click [Return] to exit:
Next, we will select this newly formed ztr component with epidote, garnet, amphibole and clinopyroxene, and discard the remaining minerals: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
2
Select a subset group of components from the following list:
ztr,TiOx,sph,ap,ep,gt,st,and,ky,sil,amp,cpx,opx
Enter as a comma separated list of labels:
ztr,ep,gt,amp,cpx
Enter c to continue: 
1 - Apply SRD correction
2 - Subset components
3 - Amalgamate components
4 - Subset samples
c - Continue
c
Finally, select 1 (or hit [Return]) to perform the CA and display the results as a biplot: 
1 - Use Chi-square distance (CA, default)
2 - Use Chi-square distance (MDS)
3 - Use Bray-Curtis dissimilarity (MDS)

Example 10: 3-way MDS

In this tutorial, we will perform a 3-way MDS analysis of five datasets from Namibia. After selecting the 7th option from the main menu, loading the datasets is done as in Example 3. The 3-way MDS function generates two pieces of graphical output: the group configuration and subject weights:

We can save these plots as vector-editable .pdf files by selecting the eigth option of the main menu: 
Pick an option:
1 - sample size calculation
2 - plot a single dataset
3 - plot multiple datasets
4 - Minsorting
5 - MDS/PCA
6 - Procrustes analysis
7 - 3-way MDS
8 - save plots (.pdf)
9 - help
q - quit
8
Enter a name for plot 2: groupconfiguration
Enter a name for plot 3: subjectweights
Which produces groupconfiguration.pdf and subjectweights.pdf.