xvalid

Process Help

XVALID - cross-validate a set of estimation parameters

Process Name	Menu Path	Link to Command Table
XVALID	Estimate ribbon \| Variograms \| Validate	Click here

Introduction
Files, Fields and Parameters
Notes
Example

Introduction

This is a Superprocess and running it may have an effect on other Datamine files in the project. More...

The process is designed to assist in the selection of parameters for grade estimation, using the cross-validation method.

How to use

The input data file is the sample data which will later be used for estimating grades into a block model. For kriging it allows different model variograms to be tested and compared. For inverse power of distance it allows different powers to be compared.

The input to XVALID is consistent with the input required for the grade estimation process ESTIMA. In fact the three input parameter files are identical for the two processes.

The cross-validation method works by removing each point in turn from the data file and estimating its value from the remaining data. In this way a table of actual and estimated values is created. A detailed statistical analysis is then carried out comparing the actuals and estimates. One or more of the estimation parameters can then be changed and the process rerun to see whether the new parameters improve the results of the statistical analysis. The method is therefore iterative, requiring several runs to establish the best set of parameters.

XVALID uses the block model grade estimation process ESTIMA to do the actual interpolation. Therefore part of the text on the Output window will be the same as for ESTIMA, and will refer to a block model rather than points. However the number of discretisation points is set to 1x1x1 so in effect point estimation is being used.

There are three input parameter files all of which can be set up using the ESTIMATE process or the Table Editor:

SRCPARM: Search Volume Parameter File
ESTPARM: Estimation Parameter File
VMODPARM: Variogram Model Parameter File

and two other input files:

IN: Sample File
VGRAM: Experimental Variogram File

The results of the cross-validation are written to three separate output files, all of which are optional:

XVSAMPS: cross-validated output sample file
XVSTATS: cross-validation statistics
SAMPOUT: sample output file containing weights for each estimate

The statistics are also displayed in the Output window as illustrated in the example. When the first set of results have been calculated and displayed you are then able to select from a menu which allows you to edit one or more of the parameter files or examine the results, and then rerun the cross-validation:

1 - edit search volume parameter file spar10

2 - edit estimation parameter file epar10

3 - edit variogram model parameter file vpar10

4 - examine variogram model interactively (VARFIT)

5 - examine cross-validation statistics file xxstats

6 - delete cross-validation statistics file xxstats

7 - create plot of actual v estimate

8 - re-run cross-validation

0 - exit cross-validation

Option 4 allows you to use the interactive variogram fitting process VARFIT. In order to use this option you must have specified the experimental variogram file VGRAM when selecting XVALID.

Options 5 and 6 are only available if the output file XVSTATS has been specified.

Option 7 is not available if multiple fields are being cross-validated ie if there is more than one record in the estimation parameter file.

Files, Fields and Parameters

Input Files

Name	Description	I/O Status	Required	Type
IN	Input sample data to be cross validated. This must contain X,Y and Z fields and at least one grade field.	Input	Yes	Undefined
SRCPARM	Search volume parameter file. This contains 24 compulsory fields defining the search volume and the number of samples needed for grade estimation. More than one search volume may be defined. All fields are numeric: SREFNUM Search volume reference number. SMETHOD Search volume shape. 1 = 3D rectangle 2 = ellipsoid. SDIST1 Max search distance in direction 1. SDIST2 Max search distance in direction 2. SDIST3 Max search distance in direction 3. SANGLE1 First rotation angle for search vol. SANGLE2 Second rotation angle. SANGLE3 Third rotation angle. SAXIS1 Axis for 1st rotation (1=X,2=Y,3=Z). SAXIS2 Axis for 2nd rotation (1=X,2=Y,3=Z). SAXIS3 Axis for 3rd rotation (1=X,2=Y,3=Z). MINNUM1 Min number of samples, 1st search vol. MAXNUM1 Max number of samples, 1st search vol. SVOLFAC2 Axis multiplying factor,2nd search vol. MINNUM2 Min number of samples, 2nd search vol. MAXNUM2 Max number of samples, 2nd search vol. SVOLFAC3 Axis multiplying factor,3rd search vol. MINNUM3 Min number of samples, 3rd search vol. MAXNUM3 Max number of samples, 3rd search vol. OCTMETH Octant method flag. 0 = no octant search, 1 = use octants. MINOCT Minimum number of octants to be filled. MINPEROC Minimum number of samples in an octant. MAXPEROC Maximum number of samples in an octant. MAXKEY Maximum number of samples with the same key value within an octant.	Input	Yes	Undefined
ESTPARM	Estimation parameter file. Each record in the file describes an estimation method and its associated parameters. The fields are dependent on the estimation methods selected. General fields: VALUE_IN: Field to be estimated. SREFNUM: Search volume reference number. VALUE_OU: Field to be created in MODEL (Default is VALUE_IN). {ZONE1_F}: 1st field for zonal estimation. The actual name of the field is given by ZONE1_F on command line e.g. ZONE1_F(ROCK). {ZONE2_F}: 2nd field for zonal estimation. NUMSAM_F: Field to be created in MODEL for the number of samples. SVOL_F: Field to be created in MODEL for dynamic search volume number. VAR_F: Field to be created in MODEL for variance of estimate. MINDIS_F: Field to be created in MODEL for distance to nearest sample. IMETHOD N Estimation method. 1 = Nearest neighbour (NN). 2 = Inverse power of dist (IPD). 3 = Ordinary kriging (OK). 4 = Simple kriging (SK). 5 = Sichel's t estimator. Anisotropy fields for NN and IPD: ANISO: Anisotropy method: 0 = no anisotropy. 1 = use search vol anisotropy. 2 = use ANANGLEn. ANANGLE1: Anisotropy angle 1. ANANGLE2: Anisotropy angle 2. ANANGLE3: Anisotropy angle 3. ANDIST1: Anisotropy distance 1. ANDIST2: Anisotropy distance 2. ANDIST3: Anisotropy distance 3. Fields for IPD: POWER Power of distance for weighting. ADDCON Constant added to distance. Fields for kriging: VREFNUM: Variogram model reference number. LOG: Lognormal variogram flag. 0 = normal kriging. 1 = lognormal kriging. KRIGNEGW: Treatment of -ve weights: 0 = -ve weights kept and used. 1 = ignore samples with -ve weights KRIGVARS Treatment of variance > sill: 0 = write variance to MODEL. 1 = set variance to sill. Fields for lognormal kriging: GENCASE Calculation method: 0 = Rendu's method. 1 = General case. DEPMEAN Deposit mean [If 0 then use kriged estimate]. Fields for general case: TOL N Tolerance for convergence. MAXITER N Maximum number of iterations. Fields for simple kriging: LOCALMNP: Method for calculation of local mean: 1 = use field defined in PROTO 2 = use mean within search vol. LOCALM_F: Name of local mean field in PROTO; used if LOCALMNP=1	Input	Yes	Undefined
VMODPARM	Variogram model parameter file. Each record in this file defines a variogram model type and its parameters. Only the VREFNUM field is compulsory. VREFNUM Model variogram reference number. VANGLE1 Variogram anisotropy angle 1. VANGLE2 Variogram anisotropy angle 2. VANGLE3 Variogram anisotropy angle 3. VAXIS1 Model variogram rotation axis 1. VAXIS2 Model variogram rotation axis 2. VAXIS3 Model variogram rotation axis 3. NUGGET Nugget variance. ST1 Variogram model type for structure 1. 1 = Spherical. 2 = Power [eg 1 - linear]. 3 = Exponential. 4 = Gaussian. 5 = De Wijsian. ST1PAR1 1st parameter of structure 1 [Range 1 for spherical model]. ST1PAR2 2nd parameter of structure 1 [Range 2 for spherical model]. ST1PAR3 3rd parameter of structure 1 [Range 3 for spherical model]. ST1PAR4 4th parameter of structure 1 [C variance for spherical model]. STn Variogram model type for structure n. STnPARp pth parameter for structure n, where n<=9.	Input	No	Variogram - Model
VGRAM	Experimental variogram file, as created by the variogram calculation process VGRAM. This experimental variogram file will have been used by the variogram fitting process VARFIT in order to derive the variogram model defined by VMODPARM . This is only required if you want to use access the variogram display and fitting process VARFIT from within XVALID.	Input	No	Variogram - Experimental

Output Files

Name	I/O Status	Required	Type	Description
XVSAMPS	Output	No	Undefined	Cross-validated output sample file. This contains all the fields from the IN sample data file, plus each grade estimate and associated secondary fields such as kriged variance.
XVSTATS	Output	No	Undefined	Output file containing a summary of the input parameters and the cross-validation statistics. It includes a single record for each estimate. The 23 fields in the file are summarised below. If the file already contains all 23 fields then additional records are appended to the file. If the file does not contain all 23 fields, or if the file does not exist, then a new file is created. ID - each estimate is assigned a unique numeric identifier VALUE_IN - field in sample IN file being estimated VALUE_OU - field containing estimated value IMETHOD - estimation method (1=NN, 2=IPD, 3=OK, 4= SK, 5=ST) SREFNUM - search volume reference number VREFNUM - variogram reference number POWER - power of distance for IPD NUM_EST - number of samples for which estimates have been made NUM_MISS - number of samples whose values were not estimated ACT_MEAN - mean grade of the sample values (the actuals) EST_MEAN - mean grade of the estimated values DIFF - actual mean grade minus estimated mean grade PC_DIFF - DIFF as a percentage of the actual mean grade MAD - mean absolute difference between actual and estimated grades ACT_VAR - variance of the actual grades EST_VAR - variance of the estimated grades COR_COEF - correlation coefficient between actuals and estimates KV_VMOD - mean of the kriged variance of each estimate KV_DIFF2 - mean of (Actual - Estimate)**2 KV_RATIO - ratio of KV_VMOD to KV_DIFF2 REG_CON - constant of regression line of actual on estimate REG_SLP - slope of regression line of actual on estimate REG_SE - standard error of regression line of actual on estimate
SAMPOUT	Output	No	Undefined	Output sample file containing details of weights for each sample for each estimate.

Fields

Name	Description	Source	Required	Type	Default
X	X coordinate of sample data in IN file. If not specified, then X is assumed.	IN	Yes	Numeric	Undefined
Y	Y coordinate of sample data in IN file. If not specified, then Y is assumed.	IN	Yes	Numeric	Undefined
Z	Z coordinate of sample data in IN file. If not specified, then Z is assumed.	IN	Yes	Numeric	Undefined
ZONE1_F	First field for zonal control. The field must exist in the IN file and in the ESTPARM file.	IN, ESTPARM	No	Any	Undefined
ZONE2_F	Second field for zonal control. The field must exist in the IN file and in the ESTPARM file.	IN, ESTPARM	No	Any	Undefined
KEY	Key field used to specify the field limiting the number of samples for estimation. The field must exist in the IN file.	IN	No	Numeric	Undefined
LENGTH_F	Field used for length weighting in IPD. The field must exist in the IN file.	IN	No	Numeric	Undefined
DENS_F	Field used for density weighting in IPD. The field must exist in the IN file.	IN	No	Numeric	Undefined

Parameters

Name

Description

Required

Default

Range

Values

SMINFAC

Multiplying factor which is applied to the first search volume, and used to calculate the exclusion volume for estimation. Samples lying within the exclusion volume are not used for the estimation. The factor must be greater than 0 and less than 1. The exclusion volume is concentric with the search volume.

0.0001

0,1

Undefined

Display control:

Option	Description
0	Minimum output.
1	Maximum output.

0,1

Notes

No additional notes.

Example

!XVALID &IN(holes.d), &SRCPARM(spar10), &ESTPARM(epar10), &VMODPARM(vpar10),

&VGRAM(vgram),&XVSAMPS(xvsamps),&XVSTATS(xvstats),&SAMPOUT(sampout),

@SMINFAC=0.00001,@PRINT=0,@ECHO=0

ESTIMA TIME >17:42:49

ESTIMA - Grade Estimation

-------------------------

Initialisation: Checking input files .............................

Initialisation: Checking estimation parameter file ...............

Initialisation: Creating estimation table ........................

Initialisation: Checking model file ..............................

Initialisation: Initialising sample output file ..................

Initialisation: Creating virtual files ...........................

Initialisation: Storing data in memory ...........................

>>> Warning - kriging has been selected and the number of

discretisation points is 1x1x1

Therefore point kriging will be used.

Initialisation: Completed.

Estimation Table

----------------

Estimations will be carried out for the following combinations of grade and

zone fields:

Sample Output Zones ................................. Search Est

Data Model Vol.Ref Meth

Grade Grade

1 AU ESTIMATE ....................................... 1.0 3

Number of records in the output model = 1077

Number of different grade fields = 1

Maximum number of estimates = 1077

This maximum number ignores retrieval criteria, selective updating, unestimated

zones etc, and so the % figure in the progress report may be too low.

>>> 10 estimates, 0.9% completed. Time 17:42:49 <<<

>>> 910 estimates, 84.5% completed. Time 17:42:54 <<<

Number of records in the output model = 1077

Number of different grade fields = 1

Maximum number of estimates = 1077

This maximum number ignores retrieval criteria, selective updating, unestimated

zones etc, and so the % figure in the progress report may be too low.

>>> 1077 estimates, 100.0% completed. Time 17:42:54 <<<

Total number of estimates 1077

Summary Statistics for Kriging

------------------------------

The total number of kriged estimates calculated is 1077

The number of kriged estimates with:

- one or more samples with zero covariance 0

- error in solving kriging matrix 0

- negative kriging variance 0

- kriging variance greater than the sill 0

- one or more negative kriging weights 0

- only one discretisation point 0

- maximum iterations in log kriging 0

>>> 1077 RECORDS IN FILE _SP15 <<<

>>> 5385 RECORDS IN FILE SAMPOUT <<<

.. calculating statistics

_____________________________________________________________________________

SUMMARY OF INPUT PARAMETERS

---------------------------

Input sample data grade field = AU

Field containing estimated values = ESTIMATE

Minimum number of samples = 5

Maximum number of samples = 15

Estimation Method: Ordinary Kriging (IMETHOD=3)

Model variogram reference number = 1

Variogram rotations are:

1st rotation around Z axis (VAXIS1) of 30 degrees (VANGLE1)

2nd rotation around X axis (VAXIS2) of 20 degrees (VANGLE2)

3rd rotation around Z axis (VAXIS3) of 0 degrees (VANGLE3)

Number of structures = 1

Nugget variance (NUGGET) = 2.5

1st Structure:

Type (ST1) = 1 (spherical)

Range in X (ST1PAR1) = 50

Range in Y (ST1PAR2) = 90

Range in Z (ST1PAR3) = 30

Spatial variance (ST1PAR4) = 9.8

Search Volume:

Search distance in X (SDIST1) = 25

Search distance in Y (SDIST2) = 45

Search distance in Z (SDIST3) = 15

Search volume rotations are:

1st rotation around Z axis (SAXIS1) of 30 degrees (SANGLE1)

2nd rotation around X axis (SAXIS2) of 20 degrees (SANGLE2)

3rd rotation around Z axis (SAXIS3) of 0 degrees (SANGLE3)

The minimum search volume (the exclusion volume) is calculated as the

volume defined above multiplied by a factor of 0.00001

__________________________________________

CROSS-VALIDATION STATISTICS FOR AU

----------------------------------------

Number of samples estimated = 1077

Number of samples not estimated = 0

Mean of actual values = 5.994147

Mean of estimated values = 5.964446

Mean difference (act - est) = 0.029706

Mean difference (as % of actual) = 0.495

Mean absolute difference = 1.612181

Variance of actual values = 7.73888

Variance of estimated values = 4.355755

Correlation coefficient = 0.666

Kriging Variance:

Mean of KV estimated from model = 4.313843

Mean of squared differences = 4.354414

Ratio = 0.99

Regression Equation:

Actual = 0.694132 + 0.888602 * Estimate

Standard Error = 2.073525

_____________________________________________________________________________

CROSS-VALIDATION MENU

---------------------

Select from: