Mars Final Project Report
+Doña Roberts
+December 12th, 2024
+ +1 DAR Project and Group +Members
+-
+
- Project name: MARS +
- Project team members:
+
-
+
- Ashton Compton
+
-
+
- dar-compta +
+ - Aadi Lahiri
+
-
+
- dar-lahira +
+ - CJ Marino
+
-
+
- dar-marinc8 +
+ - Nicolas Morawski
+
-
+
- dar-morawn +
+ - Dante Mwatibo
+
-
+
- dar-mwatid +
+ - Charlotte Peterson
+
-
+
- dar-peterc +
+ - Doña Roberts
+
-
+
- dar-roberd10 +
+ - Margo VanEsselstyn
+
-
+
- dar-vanesm +
+ - David Walczyk
+
-
+
- dar-walczd3 +
+
+ - Ashton Compton
+
2 0.0 Preliminaries.
+This report is generated from an R Markdown file that includes all
+the R code necessary to produce the results described and embedded in
+the report. Code blocks can be suppressed from output for readability.
+If show <- FALSE the code block will be suppressed; if
+show <- TRUE then the code will be shown.
# Set to TRUE to expand R code blocks; set to FALSE to collapse R code blocks
+show <- TRUEExecuting this R notebook requires the following packages:
+-
+
ggplot2
+tidyverse
+pheatmap
+reshape2
+
These will be installed and loaded as necessary.
+3 1.0 Project +Introduction
+This report is for the Mars 2024 Data Analytics Research group. Our +goal is to create an app (“Mission Minder”) that shows unique dynamic +analysis of the Mars Perseverance data.
+This particular report mainly focuses on how the data analyzed in +“Mission Minder” is cleaned and organized. With a brief look at how the +two types of features in sample data (mineral presence and elemental +compound concentration) relate and why we won’t be using Lithology going +forward.
+4 2.0 Organization of +Report
+This report is organized as follows:
+-
+
Section 3.0. “Finding 1: Data Organization”
+-
+
- Details of how we reformatted and cleaned the data for consistency +and ease of use over multiple analysis methods. +
- Notes on decisions made in the cleaning process, their consequences, +and potential alternative choices. +
- Instructions on how to integrate new samples/targets into the data +sets. +
+Section 4.0: “Finding 2: Correlation Between Elemental Compounds +and Minerals”
+-
+
- Calculating the correlation between the two types of numeric +features (PIXL & SHERLOC) of the sample data. +
+Section 5.0: “Lithology and SHERLOC match”
+-
+
- Showing that Lithology is just the binary representation of +SHERLOC. +
+Section 6.0 “Overall conclusions and suggestions”
+Section 7.0 “Bibliography”
+Section 8.0 “Appendix”
+-
+
- Instructions on how to convert Lithology/SHERLOC data from factor to +the accurate numeric. +
+
5 3.0 Finding 1: Data +Organization
+The original data for Perseverance was inconsistently organized and +not always in a usable form. This section describes how the data was +reorganized, re-classed, and relabeled to be more intuitive and to +minimize the data manipulation needed for each individual analysis.
+5.1 3.1 Data, Code, and +Resources
+Here is a list of data sets, codes, and resources that relate to this +section. Note that if you click on the file names it will send you to +that file in the GitHub.
+-
+
- Code altering the data sets
+
-
+
- v1_consistent_data_naming.Rmd +is the notebook that cleans the data and outputs all the v1 data sets. +All the cleaning done to the code is described in the notebook and +should be easily editable. +
+ - Markdowns describing data sets
+
-
+
- v1_Data_Introduction.md +is the markdown file describing each of the new v1 data sets, what they +contain, and how they have been changed. +
- “Dataset-Description”
+is the GitHub wiki page describing each of the v1 data sets. As of
+December 14th, 2024, this is more up to date than
+
v1_Data_Introuduction.md.
+
+ - Data sets
+
-
+
- v1_sample_meta.Rds +is the data set containing all the meta data for the physical samples +Perseverance collected. +
- v1_pixl.Rds +is the data set containing all the elemental compound concentration data +for the physical samples Perseverance collected. +
- v1_sherloc.Rds +is the data set indicating confidence in mineral presence (NOT bindary) +in the physical samples Perseverance collected. +
- v1_lithology.Rds +is the data set indicating whether minerals were present or not present +(binary) in the physical samples Perseverance collected. +
- v1_libs.Rds +is the data set containing all the meta data and the elemental compound +concentration data for the LIBS scans Perseverance did on Mars +rock. +
- v1_libs_to_sample.Rds +is the data set containing the nearest sample to each LIBS target/point +and the distance between them. +
- v1_libs_earth_references.Rds +is the data set containing all the meta data and the elemental compound +concentration data for the LIBS scans Perseverance did on the Earth +reference rocks it carries with it. That is, the “scct” targets. +
+ - Final
+Notebooks
+
-
+
- peterc_finalProjectF24.Rmd +is Charlotte’s final notebook. Referenced in part 3.2, 3.3, and +3.5. +
- vanesm_finalProjectF24.Rmd +is Margo’s final notebook. Referenced in part 3.2, 3.3, and 3.5. +
- walczd3_finalProjectF24.Rmd +is David’s final notebook. Referenced in part 3.2. +
+
5.2 3.2 Contribution
+The reordering and data cleaning was done by me, Doña Roberts.
+The data sets v1_libs_to_sample.Rds, the Type meta data
+for v1_libs.Rds, and the Sol, Lat, and Lon meta data for
+v1_pixl.Rds came from date sets created by Margo
+VanEsselstyn and Charlotte Peterson. Their final notebooks, which
+document where they got the new data, can be found on the GitHub
+following links 3.1.d.1 and 3.1.d.2.
Additionally, David created a data set, aqueous.Rds
+that describes the details of each of the minerals in SHERLOC. His final
+notebook, which documents how he created this data set, can also be
+found on GitHub following link 3.1.d.3.
5.3 3.3 Data +Description
+For the description of the resulting data sets, see the wiki page: https://github.rpi.edu/DataINCITE/DAR-Mars-F24/wiki/Dataset-Description
+Since the wiki doesn’t go into detail on what changes were made (just
+what the end result is) we’ll summarize that now. Note that the actual
+file implementing these changes (with comments) is
+v1_consistent_data_naming.Rmd and can be found by on GitHub
+by following link 3.1.a.1.
-
+
- Imported data sets. These are the data I started with and then
+manipulated to create the v1s.
+
-
+
- pixl_sol_coordinates.Rds, +which is the PIXL data set including Sol/Lat/Lon created by Charlotte +and Margo. See their final notebooks for how they created this +file. +
- supercam_libs_moc_loc.Rds, +which is the LIBS data. This comes from a csv downloaded from NASA. +
- libs_typed.Rds, +which is the LIBS data set including a column stating what type of scan +it was. This was also created by Charlotte and Margo. See their final +notebooks for how they created this file. +
- mineral_data_static.Rds, +which is the Lithology data. This comes from the “Lithology” paragraph +at the start of each sample’s sample +report. +
- abrasions_sherloc_samples.Rds,
+which is the SHERLOC data set that comes from the “SHERLOC” tables in
+the sample
+reports.
+
-
+
- Some manipulation was done to this data set to make it’s format +match the others’. +
+ - PIXL_LIBS_Combined.Rds, +which is the data set listing each LIBS targets’ nearest abrasion and +the distance to it. This was created by Charlotte and Margo. See their +final notebooks for how they created this file. +
+ - Renamed Columns
+
-
+
- Capitalized all column names for consistency +
- Changed PIXL elemental compound naming convention to match +LIBS’s +
- Created a logical vector from the “earthsample?” column of
+
libs_typed.Rds(to be used later in separating out the +earth references)
+ - Changed Lithology & SHERLOC’s mineral naming convention to be +consistent between minerals and with each other +
- Changed names of
PIXL_LIBS_Combined.Rds(now called +“libs_to_sample”) so that it matches the naming convention of PIXL/LIBS +and to make explicit whether a column came from PIXL or came from +LIBS
+
+ - Separated out Sample Meta data
+
-
+
- See section 3.4’s table of decisions, row 2, for details. +
+ - Separated out LIBS earth reference data from rest of data
+
-
+
- See section 3.4’s table of decisions, row 3, for details. +
+ - Created the LIBS clusters used in most LIBS analysis and creating a
+column in the LIBS data frame indicating said clustering
+
-
+
- See section 3.4’s table of decisions, row 6, for details. +
+ - Reordered Columns
+
-
+
- Reordered LIBS columns such that the meta data is all at the +front +
- Reordered PIXL columns such that the elemental compounds also found +in LIBS data are ordered the same, with elemental compound data unique +to PIXL is at the end +
- Reordered Lithology and Sherloc so that they match +
- Reordered
libs_to_samplecolumns such that LIBS meta +data was first, distance between, PIXL meta data next, and LIBS +elemental compound data last. +-
+
- See section 3.4’s table of decisions, row 4, for details on the +inclusion of the LIBS elemental compound data and all of the meta +data. +
+
+ - Changed data types
+
-
+
- New data types are listed in the Wiki linked at the start of this +section. +
- Several
charactertypes were changed to +factor+-
+
- See section 3.4’s table of decisions, row 1, for details on one such +change. +
+ - Sample Meta was originally all character, numbers were changed to
+
numericorintegerand labels were changed to +factor
+
+ - Saved as new Rds files
+
-
+
- All new Rds file names are proceeded with
v1_with the +intention that updates to these data sets will be saved with the prefix +vn_where “n” denotes what addition they are. This way, if +any errors occur in the updating process, the Mission Minder can still +run on the older data sets while the new ones are fixed.
+
+ - All new Rds file names are proceeded with
5.4 3.4 Limitations
+While I attempted to organize the data in the most useful ways, there +were some changes that could have been un-ideal, inconvenient, or had a +valid alternative.
+Here is a table of the most notable ones:
+Table of Decisions
+| File: | +Part in question: | +Decision: | +Pros: | +Cons: | +Alternative: | +
|---|---|---|---|---|---|
v1_sherloc |
+The class of mineral presences | +Setting the class as factor |
+Accurate to what it is | +Inconvenient to use, see 8.1 for how to convert to numeric | +1) Leaving class as character 2) Setting class as
+numeric |
+
| _________ | +_______________ | +_________ | +__________ | +__________ | +______________ | +
v1_sample_ meta |
+The existence of a separate meta data set | +Separating meta data out | +Cleaner, Neater, Less repetitive | +One more data set to import | +Repeating meta data in each sample data set | +
| _________ | +_______________ | +_________ | +__________ | +__________ | +______________ | +
v1_libs_earth _references |
+The existence of a separate earth reference data set | +Separating scct (Earth reference) targets out |
+We aren’t accidentally looking at Earth rocks when trying to analyze +Mars ones | +One more data set to import, harder to compare scct
+data to other types of libs data |
+Keeping scct targets in the v1_libs data
+set |
+
| _________ | +_______________ | +_________ | +__________ | +__________ | +______________ | +
v1_libs_to_ sample |
+The inclusion of non-identifier PIXL and LIBS data | +Include the extra meta and elemental compound data | +Convenience. Change would be too late in process, you’d have to go +through and change any code depending on the “fluff’s” inclusion | +Makes the data set more complex, more repetitive, and less +clean | +Removing non-identifier data from
+v1_libs_to_sample |
+
| _________ | +_______________ | +_________ | +__________ | +__________ | +______________ | +
v1_consistent _data_naming |
+Line 153, the inclusion of all samples in
+v1_lithology |
+Limiting Lithology to only first 16 samples (by several +requests) | +Don’t need to remove partially received samples for every analysis +including PIXL data | +Problematic for when we receive new samples!!! | +1) Don’t limit it but people need to manually remove extra rows for +analysis 2) Define a variable “n” at start of file, limit to “n” +samples. Increment “n” every time we get a new sample | +
| _________ | +_______________ | +_________ | +__________ | +__________ | +______________ | +
v1_libs |
+The inclusion of a “Cluster” column | +Include cluster column | +For analysis using the 4 clusters, it’s easier. Unifies clustering +between people’s analysis | +Doesn’t really make sense with the app where number of clusters is +changeable and seed is already consistent between analysis | +Don’t include cluster column, instead rely on a “Cluster” variable +in the app which is dependent on the number of clusters requested | +
There are many more decisions made while cleaning the data, but the +above table details the most important (questionable) ones.
+5.5 3.5 Future Work
+In the future, we expect to receive more Sample and LIBS data from +NASA. When this happens, there needs to be a way to integrate it.
+Below are the steps required, and the files to reference, in order to +add and clean the new data. For some of the steps the code to complete +it isn’t public in the Data or StudentData GitHub folders, so the person +responsible for that code or the place to find it detailed is +identified.
+For new Sample data:
+-
+
Get the new data from NASA to update +
+mineral_data_static.Rds, +abrasions_sherloc_samples.Rds, and +sample_pixl_wide.Rds(which is used in the creation of +pixl_sol_coordinates.Rds)-
+
- For updating
mineral_data_static.Rds, look at the +“Lithology” report at the start of each new sample’s Sample +Report. As of December 14th 2024, the specific code to create the +Rds file isn’t public, so contact Dr. Erickson.
+ - For updating
abrasions_sherloc_samples.Rds, look at the +“SHERLOC” table in each new sample’s Sample +Report. Talk to Karen Rogers for details.
+ - For updating
sample_pixl_wide.Rds, as of December 14th +2024, there is no documentation available on GitHub explaining how this +Rdsfile was created. Check with Dr. Erickson to see where +this data came from and how to update it.
+
+- For updating
Follow instructions in Margo and Charlotte’s notebooks to update +
pixl_sol_coordinates.Rds. The Sol/Lat/Lon information they +added comes from the Analyst +Notebook.
+Follow instructions in Margo and Charlotte’s notebooks to update +
PIXL_LIBS_Combined.Rdsfor the new LIBS points.
+Change line 153 of
v1_consistent_data_naming.Rmd+fromlithology.df <- lithology.df[1:16,~~~]to +lithology.df <- lithology.df[1:n,~~~]where +nis the new number of samples.
+Run
+v1_consistent_data_naming.Rmdto update +v1_sample_meta.Rds,v1_pixl.Rds, +v1_sherloc.Rds, andv1_libs_to_sample.Rds.-
+
- Note that since Lithology has become redundant, there is no need to
+update
v1_lithology.Rds.
+
+- Note that since Lithology has become redundant, there is no need to
+update
For new LIBS data:
+-
+
Download new moc csv file from LIBS, and update +
supercam_libs_moc_loc.Rds. As of December 14th 2024, the +specific code to create the Rds file isn’t public, so contact +Dr. Erickson.
+Follow instructions in Margo and Charlotte’s notebooks to create +
libs_typed.Rdsfor the new LIBS points.
+Follow instructions in Margo and Charlotte’s notebooks to update +
PIXL_LIBS_Combined.Rdsfor the new LIBS points.
+Run
v1_consistent_data_naming.Rmdto update +v1_libs.Rds,v1_libs_earth_references.Rds, and +v1_libs_to_sample.Rds.
+
Some recommendations for updating the data sets:
+-
+
Instead of updating “v1” to contain the new data, create a copy +of
v1_consistent_data_naming.Rmdand replace all references +to “v1” with “v2” (including the title of the file). This way, instead +of updating the “v1” data sets you are creating new “v2” data sets. This +means that if something goes wrong and it breaks, you can always fall +back to the previous version while you figure out how to fix the +break.
+When creating the
v2_consistent_data_naming.Rmd, go +through and find Margo and Charlotte’s code to update their data sets +and add it to the start of the newv2code so that in the +future there is only one file to run when updating. I was going to do +this but, as they were still improving and updating it, the code wasn’t +stable enough for the transfer to make sense.
+
6 4.0 Finding 2: +Correlation Between Elemental Compounds and Minerals
+This section looks at how the concentration of elemental compounds +(PIXL data) relates to/indicates the presence of minerals (SHERLOC data) +at a sample location.
+We are looking at the question of how PIXL and SHERLOC relate to each +other, and thus the [Exploration] –> [PIXL vs Sherloc] section of the +Mission Minder.
+6.1 4.1 Data, Code, and +Resources
+Here is a list of data sets, codes, and resources that are used in +this work. Click on the file name to be sent to the file on GitHub.
+-
+
- Data sets (code for importation is hidden)
+
-
+
- v1_pixl.Rds +
- v1_sherloc.Rds +with the exception of “Hydrated Carbonates” which is the same for all +samples. +
+
First, we need to report the data sets.
+Then, before we can actually calculate the data we need, we must +combine the two data sets into a matrix with only numeric data.
+## Creating pixl matrix
+# removing the sample number column
+pixl.matrix <- as.matrix(pixl.df[,-1])
+## Creating sherloc matrix
+# removing the sample number column
+# converting from feature to numeric
+sherloc.matrix <- as.matrix(as.data.frame(lapply(lapply(
+ sherloc.df[,-1],as.character),as.numeric)))
+## Combining pixl and sherloc into one matrix
+# removing the column(s) with no deviation, in this case "Hydrated Carbonates"
+combined.matrix <- cbind(pixl.matrix,sherloc.matrix[,-16])With the data from the combined matrix, we calculate the pearson +correlation of their combined data, and then look at only the +correlation between features in PIXL and features in SHERLOC
+# Calculating pearson feature correlation
+combined.cormat <- round(cor(combined.matrix),2)
+# Selecting only the correlations between pixl features and sherloc features
+combined.cormat <- combined.cormat[colnames(pixl.matrix),colnames(sherloc.matrix[,-16])]6.2 4.2 Contribution
+This section is solely my own work.
+6.3 4.3 Methods +Description
+In order to get from raw SHERLOC and PIXL data to their correlation,
+I simply used the cor function from base r to calculate the
+correlation between each feature in PIXL (elemental compounds) and
+SHERLOC (minerals) and then selected only the correlations between
+features of PIXL and features of SHERLOC (instead of correlations within
+PIXL or within SHERLOC) to find my desired correlations.
In the next section, I simply use pheatmap to
+hierarchaly cluster and output a heatmap of correlations.
6.4 4.4 Result and +Discussion
+# Creating title for heatmap
+heatmap.title <- "Correlation between Elemental Concentration & Mineral Presence"
+# Printing heatmap of correlation
+pheatmap(combined.cormat,
+ scale="none",
+ treeheight_row = 10,treeheight_col = 10,
+ main = heatmap.title)
+Caption: Vertical axis is PIXL elemental compounds, horizontal axis is
+SHERLOC minerals. Colored squares indicate the correlation between the
+elemental compound in that row and the mineral in that column. That is
+to say, if the square is red then when that elemental compound is
+present it is highly likely that the mineral is present and if that
+square is dark blue then when that elemental compound is present it is
+highly unlikely that the mineral is present.
Here we can clearly see see which elemental compounds correlate to +which minerals.
+For example, high SO3 concentration strongly correlates to the +presence of Hydrated Mg-Fe sulfate, Mg-sulfate, Kaolinite, & +Fe-Mg-clay minerals.
+-
+
- Here the sulfates make sense, as sulfate is SO4, or, O + +SO3 +
- Interestingly though, SO3 doesn’t have a strong +correlation with Sulfate itself. +
Similarly, high Cr2O3 concentration strongly correlates to the +presence of Apatite, Spinels, Zircon/Baddeleyite, Chromite, & +Ilmenite.
+-
+
- Here Chromite makes sense, since it’s FeCr2O4, or, FeO + +Cr2O3 +
Overall we see that;
+-
+
- MgO: Lightly suggests presence of Fe-Mg carbonate, very +lightly correlated to a bunch of minerals but not strongly correlated to +the presence of any of the minerals +
- Cr2O3: Strongly suggests the presence of Apatite, +Spinels, Zircon/Baddeleyite, Chromite, & Ilmenite. +
- Cl: Strongly suggests the presence of Fe-Mg +carbonate, & Halite. Very lightly correlated to a few other +minerals. +
- MnO: Mildly suggests the presence of Carbonate and lightly corelated +to presence of Pyroxene. Very lightly correlated to a bunch of other +minerals, but nothing strong. +
- FeO-T: Mildly suggests the presence of Carbonate, and very lightly +correlated to a bunch of minerals. +
- SO3: Strongly suggests the presence of Hydrated +Mg-Fe sulfate, Mg-sulfate, Kaolinite, & Fe-Mg-clay minerals. Lightly +suggests the presence of Ca-Sulfate. +
- TiO2: Lightly suggests presence of Plagioclase, and very lightly +suggests the presence of several minerals. Again, no strong correlations +though. +
- P2O5: Mildly suggests the presence of Plagioclase, Iron oxide, +Ca-sulfate, & FeTi oxides. Lightly suggests several other +minerals. +
- CaO: Mildly suggests the presence of Hydrated Iron oxide, +Perchlorates, Na-perchlorate, Plagioclase, & Iron oxide. A bunch of +light correlations to other minerals as well. +
- Na2O: Mildly suggests the presence Iron oxide. Lightl suggests the +presence of a bunch of other minerals as well. +
- Al2O3: Mildly suggests the presence of Chlorite, Quartz, Disordered +Silicates, & Feldspar. Very lightly suggests the presence of a bunch +of other minerals. +
- SiO2: Mildly suggests the presence of Chlorite, Quartz, Disordered +Silicates, & Feldspar. Very lightly suggests the presence of several +other minerals. +
- K2O: Strongly suggests the presence of Chlorite, Quartz, Disordered +Silicates, & Feldspar. Very lightly suggests the presence of a few +other minerals. +
6.5 4.5 Conclusions, +Limitations, and Future Work.
+There is likely a chemical reason for the strong correlations found +above. Some correlations are obvious even to a non geologist, such as +Chromite & Cr2O3 or several Sulfates & SO3, but other +correlations aren’t immediately obvious, such as the low correlation +between Sulfate & SO3 or Perchlorates & Cl. There is likely a +lot more connections to be seen here by someone with more geological +insight.
+It would also be interesting to look at this heatmap in conjunction
+with the chemical formulas for the minerals listed in David’s aqueous.Rds
+file and see if that provides any new insights.
Sadly, since, we don’t have data for mineral presence at LIBS +targets, we can’t do a similar heatmap for that, but we may be able to +look at the relation between concentration of elemental compounds and +presence of minerals found through sample data to hypothesize what +minerals are likely present at the LIBS targets based on the elemental +compound concentrations found there.
+7 5.0 Finding 3: +Lithology and SHERLOC match
+This section looks at the realtionship between the Lithology data and +the SHERLOC data.
+We started with looking at two data sets, Lithology and SHERLOC, that +both describe the presence of minerals at each sample site.
+Lithology is made up of binary numeric values representing present or +not present, and SHERLOC is made up of discrete numeric values +representing the level of presence.
+If these two match, then we only really need SHERLOC, as Lithology +can be achieved by looking at if a SHERLOC value is “= 0” (then “0”) or +“> 0” (then “1”).
+For this reason, we want to confirm that SHERLOC and Lithology +match.
+7.1 5.1 Data, Code, and +Resources
+Here is a list of data sets, codes, and resources that are used in +this work. The data sets below can be found on GitHub by clicking on +them.
+-
+
- Data sets (code for importation is hidden) + +
First we import the Lithology data set. Note that we don’t need to +import SHERLOC, since we imported that in section 4.1.
+Then, we convert Lithology into a matrix, excluding the “Sample” +column. Again, note that we don’t need to do this for SHERLOC, as we +also did this in section 4.1.
+## Creating lithology matrix
+# removing the sample number column
+# converting from feature to numeric
+lithology.matrix <- as.matrix(as.data.frame(lapply(lapply(
+ lithology.df[,-1],as.character),as.numeric))) 7.2 5.2 Contribution
+This section is solely my own work.
+7.3 5.3 Methods +Description
+There are many ways to confirm that Lithology and SHERLOC match, but +the simplest one is probably to just subtract the SHERLOC matrix from +the Lithology matrix and look at the minimum and maximum outputs.
+## Calculating Difference matrix
+diff <- lithology.matrix - sherloc.matrix
+## Finding minimum, should be 0
+min(diff)## [1] 0## Finding maximum, should be 0.75
+max(diff)## [1] 0.75After doing this, we get that the minimum of the difference is “0” +and the maximum of the difference is “0.75”.
+7.4 5.4 Result and +Discussion
+Since the minimum of the difference between them is “0”, we know that +if Lithology is “0” then SHERLOC is also “0”, since if SHERLOC were > +“0” then difference would be < “0”.
+Since the maximum of the difference between them is “0.75”, we know +that if Lithology is “1” then SHERLOC is \(\geq\) “0.25”, since if SHERLOC were < +“0.25” then difference would be > “0.75”.
+This means that we know if Lithology claims a mineral is present then +SHERLOC also claims the mineral is present, and that if Lithology claims +that a mineral is absent then SHERLOC also claims that the mineral is +absent.
+7.5 5.5 Conclusions, +Limitations, and Future Work.
+Since we can show that SHERLOC and Lithology match, we have decided +to only use the SHERLOC data set in Mars Mission Minder, since Lithology +is simply a less detailed version of SHERLOC.
+If there are any carry over references to “Lithology” within the Mars +Mission Minder, this is actually referring to the binary version of +SHERLOC.
+8 6.0 Overall conclusions +and suggestions.
+In the future, as more samples are added, I would suggest +streamlining the process of integrating new data. That is, I would +recommend creating one master file that pulls in all the data in their +raw form (csv’s or manually added data from sample reports), cleans and +organizes it, and then outputs the finalized data sets. The current +system has many intermediary steps, some of which aren’t available for +viewing and editing, making it hard to integrate new data.
+When the Mission Minder gets extended to the other Mars Missions, I +would recommend (at the start!) sitting down and finding some uniform +naming, organizing, and updating system that can work for +ALL the missions. This way, when analysis +between missions starts to be created, the data will already +line up neatly without extra manipulation required. I would recommend +this to be a priority, done before analysis starts being created, so +that the data is ready and finalized before it starts being used. Many +of the things that make our current cleaning process and our finalized +data sets unideal were caused by people using and changing the +pre-existing (uncleaned) data.
+9 7.0 Bibliography
+-
+
- Citations from literature.
+
-
+
- “Mars Rock Samples - NASA Science.” NASA, science.nasa.gov/mission/mars-2020-perseverance/mars-rock-samples/. +Accessed 12 Dec. 2024. +
- “Analyst Notebook”, https://an.rsl.wustl.edu/m20/AN/account/login.aspx. +Note that an account is required to access the notebook. +
+ - Significant R packages used
+
-
+
- reshape2, Hadley Wickham (2007). Reshaping Data with the reshape +Package. Journal of Statistical Software, 21(12), 1-20. URL http://www.jstatsoft.org/v21/i12/. +
- pheatmap, Kolde R (2019). pheatmap: Pretty Heatmaps. R +package version 1.0.12, https://CRAN.R-project.org/package=pheatmap. +
+
10 8.0 Appendix
+10.1 8.1 Converting +Lithology/SHERLOC factor data to the correct numeric
+If you attempt to directly convert Lithology or SHERLOC from factor
+to numeric using as.factor you run into a problem. The
+“number” factor data gets changed into the wrong “number” numeric data.
+As seen in the table below.
| Factor # | +Lithology Basic Conversion | +SHERLOC Basic Conversion | +Desired Conversion | +
|---|---|---|---|
| “0” | +1 | +1 | +0 | +
| “0.25” | ++ | 2 | +0.25 | +
| “0.5” | ++ | 3 | +0.5 | +
| “0.75” | ++ | 4 | +0.75 | +
| “1” | +2 | +5 | +1 | +
The code chunk displayed below shows my method for getting around +this.
+# Converting factor "numbers" to equivalent numeric data
+lithology.df[,-1] <- as.data.frame(lapply(lapply(
+ lithology.df[,-1],as.character),as.numeric))
+sherloc.df[,-1] <- as.data.frame(lapply(lapply(
+ sherloc.df[,-1],as.character),as.numeric))Code explained:
+-
+
- We don’t change the 1st column (
[,-1]), since that is +“Sample” ID and that should remain asinteger.
+ - Each class change is done using
+
lapply(dataset,as.newclass)since we are working with a +data frame and are trying to individually change the class of each +element of the data frame. The lapply function (part of base r) is a +convenient way to do this.
+ - We first convert to
characterbefore converting to +numericbecausefactorsmoothly converts to +the identicalcharacter(I.e., if you have the string +“Blarg” as a factor, the character version will be the identical string +“Blarg”) but it doesn’t smoothly convert tonumeric(I.e., +if it is the string “Blarg” as a factor, the numeric version would be +1, assuming “Blarg” is the first factor listed. Similarily +if the string “0” is the first factor, it will convert to the numeric +“1”). Then, sincecharactersmoothly converts to +numericif the character is a +number, we can now convert without complications.
+ - We say
as.data.frame, since lapply’s normal output is a +list. This isn’t really needed here, because r interprets the +[,-1]to mean we want a dataframe not a list and will give +us that on it’s own, but for safety it’s best to include this, as there +may be another case where you are trying to convert the entire +data frame, and you’ll thus need to useas.data.frame.
+