Unraveling the Mystery of George W. Dempsey, son of Seaton Y. Dempsey and Clementine Gowing (Part 3)

George W. DEMPSEY, son of Seaton Y. DEMPSEY and Clementine M. GOWING, was born in Amherst County, Virginia, about 1831. He moved to Fayette County about 1855 before West Virginia became a state. After the 1870 census, George disappeared or died without records. He was discussed in Unraveling the Mystery of George W. Dempsey, son of Seaton Y. Dempsey and Clementine Gowing (part 1).

I hadn’t thought to investigate the whereabouts of George W. DEMPSEY, my 2nd great-granduncle until I discovered a group of DNA matches who descend from Mollie Lee DEMPSTER (1880-1950). Her story was told in Unraveling the Mystery of George W. Dempsey, son of Seaton Y. Dempsey and Clementine Gowing (part 2).

Mollie’s father was Wesley G. DEMPSTER, a man who appeared in Scott County, Virginia, shortly before the 1880 census. He likely died between 23 November 1886 and 15 December 1887. A death record was not found.

Mollie married at the age of 16 and had a family of nine children born between 1898 and 1917. Six of these children have descendants who’ve had their DNA tested. Descendants of the other three may have tested. They haven’t been found on the match lists of the tests I have access to.

Can DNA unravel the mystery of George W. Dempsey’s disappearance?

It’s complicated! I’ve been learning about DNA since the end of May 2016 when my brother turned his AncestryDNA test over to me. It has been a slow, uphill climb learning to work with the DNA results. I know this post may be hard to follow, I hope I haven’t made it too complicated. I’m assuming my readers have a basic understanding of autosomal DNA.


This is an example of one of my notes on Ancestry for a match:
[C8] 1C (Lois) Fred Rothwell DEMPSEY and Myrtle Hazel ROOP.
In brackets is the cluster number (from the first time I clustered my matches) followed by the level of cousinship. In parenthesis is the name of the child of the most recent common ancestors (MRCA) that the match descends from followed by the MRCA.

My private but searchable family tree is attached to the DNA tests I manage. Confirmed matches are connected in this tree. The tree is also used to work out unknown matches.

As I have few maternal matches and my mother has tested, all maternal matches are starred. This allows me to use all 24 colors for custom groups for my paternal matches. I created custom groups for each of my paternal 4th great-grandparent couples. The four blue colors were used a bit differently than the green, pink, and yellow as there is a brick wall at the 3rd great-grandparent level for my William A. W. DEMPSEY. He is not from the same line as Seaton Y. DEMPSEY.

16 custom color groups for the paternal 4th great-grandparent couples

Ancestral Quest’s Color Coding feature made it easy to work out the custom color groups on Ancestry.

My paternal grandfather’s pedigree.
My paternal grandmother’s pedigree.

Paternal first cousins share the DEMPSEY-ROOP couple with me and are given each of the 16 custom groups (4 shades of the 4 colors). Second cousins who share DEMPSEY-INGRAM receive 8 custom groups (4 shades of blue and of green). Third cousins who share INGRAM-DEMPSEY receive 4 custom groups (4 shades of green). This is one way to visually cluster matches.

Note: The same system can be used for both maternal and paternal matches. In this case, the 5th generation (3rd great-grandparents) is used instead of the 6th generation (4th great-grandparents) as seen in my example.

This is my top match in the group of matches who descend from Mollie on Ancestry. The top shared matches (ICW = in common with) with Match 1 are two of my first cousins with whom I share grandparents Fred R. DEMPSEY and Myrtle H. ROOP. The next two ICW matches are both 1C1R but not from the same generation. This is confirmed by the colored groups. The match with only blue and green is a 1C1R through my paternal grandfather’s parents.


I have guest or collaborator access to a few of my DEMPSEY cousins’ AncestryDNA. They have given me permission to use their tests as examples along with their first names or initials. In the image above, the two cousins with trees are the 1C1R (E.D.) and 1C (Laura) in the table below.

DNA matches descending from 6 of Mollie’s 9 children were found to match 6 tests I have access to. E.D. (1C1R) is my father’s paternal first cousin. She is a generation closer to Seaton and Clementine than myself, my brother, my first cousin Danny, and my second cousins, Laura and Sheila. The second cousins are E.D.’s nieces through two of her siblings. If they had been her children I would not have used them as they would carry the same DNA and would only duplicate the results. All of the cousins have their DNA uploaded to Gedmatch or MyHeritage except for Sheila.

Shared Clustering Tool

My brother’s and my AncestryDNA tests were clustered using Jonathan Brecher’s Shared Clustering Tool. Clustering has given me a relatively good idea of where in the family tree a match or group of matches fit in.

Jonathan’s method uses all matches and shared matches (ICW) down to 6-8 cMs on Ancestry to form clusters that point to a shared ancestor. A cluster represents a DNA segment shared by the clustered matches. Even though Ancestry does not offer a chromosome browser, the segments can be ascertained (guessed) by comparing to matches who’ve transferred their AncestryDNA to FTDNA, MyHeritage, or Gedmatch.

The data needed for clustering was downloaded from Ancestry using the Shared Clustering Tool. I’ve been manually adding new matches since Jonathan disabled downloading of data from Ancestry in May 2020. Soon after this, Ancestry sent cease and desist orders to many third-party tools.

Early this month, I subscribed to DNAGedcom for $5/month to get an up-to-date list of matches and of ICW matches from Ancestry using the DNAGedcom Client. The ICW match list can be used to generate clusters using the Shared Clustering Tool.

Screenshot of part of a cluster report generated by Shared Clustering Tool. Clusters have a blue outline and may overlap. The green highlights in this clip were added later.
Genealogical DNA Analysis Tool (GDAT)

Becky Mason Walker’s Genealogical DNA Analysis Tool (GDAT) is the repository I use to manage my DNA tests.

The database is stored locally on my computer and has no connection to the internet. I can import all DNA matches from the different testing companies, do triangulation and in common with (ICW) comparisons, map the chromosomes of common ancestors, mark the most recent common ancestors (MRCA), add Ahnentafels of the matches, and do analysis work that helps with the family tree research. With all information in one place, the tool provides easier-to-see patterns and clues to solve the genetic genealogy questions.

The Barron-Dempster matches who descend from Mollie were found to be in clusters [C54], [C29], [C30], and [C8]. All notes on Ancestry have been imported into GDAT. Since my notes begin with the cluster number, I can sort matches to view a list of only the relatives (matches) in a particular cluster.

Screenshot of GDAT Relative List sorted to show only [C54] matches with privatized names.
Cluster [C54] is large with over 400 matches ranging from 229 cMs down to 7 cMs. The identified relatives have the following MRCA: Dempsey-Ingram, Dempsey-Gowing, Going-Potter, and Crisp-Lucy. These are parents, grandparents, and great-grandparents of Mary M. DEMPSEY, daughter of Seaton Y. DEMPSEY and Clementine M. GOWING. The cluster appears to be pointing to the GOWING branch but the many matches that are still unknown will help to “walk the segment back” to the shared distant ancestor.

Of these over 400 matches, nine were found on sites with chromosome browsers. None of these have a confirmed MRCA but they share DNA on the same segment (different lengths) on chromosome 9. This segment is also shared with E.D., Danny, and Laura seen in the DNA comparison table (above, in the Ancestry section). The red segments (below) are Danny, his sister, and my Dad’s Lazarus kit. They share my paternal grandfather (PGF) and paternal grandmother’s (PGM) lines, i.e. DEMPSEY-ROOP. The blue segments are people who share only my PGM’s line, i.e. DEMPSEY-INGRAM, and include Laura and E.D.

Screenshot of GDAT Chromosome Browser information with privatized names.

Using the same process as above, I found:

  • [C29] includes about 200 matches. Only two in the cluster have chromosome data and share a segment on Chr. 6. An MRCA has not been found for either. The segment triangulates with a known 4C1R (George W.) Seaton Y. DEMPSEY and Clementine M. GOWING as well as E.D. Danny did not receive this segment but his sister (who did not test with Ancestry) is one of the matches who triangulate with the [C29] matches.
  • [C30] has about 100 matches. MRCAs in the cluster include Ingram-Dempsey(1), Dempsey-Gowing(20), Gowing-Crisp(3), Going-Potter(1), and Crisp-Lucy(4). The cluster is associated with a segment on Chr. 2 shared with E.D., Danny’s sister, and Laura.
  • [C8] has about 120 matches. This is E.D., Danny, and Laura’s cluster. They correlate with many other clusters but this is their main cluster. MRCAs in the cluster include Dempsey-Wood, Wood-Honaker, Wood-McGraw which suggest the cluster is coming from the PGF (blue) side. The two Barron-Dempster matches (Match 2 and 5, father and daughter) associated with this cluster share at two segments with several of us. One of these segments may have a distant connection to the blue side.

My brother received very little DNA shared with the Barron-Dempster matches – only a 12 cMs segment with Match 1 and 9 cMs of the same segment with Match 1a (child of 1).

Shared Clustering

Clusters fluctuate as new matches are added. Since my test was clustered in September 2019 many new matches have been added. I ran a new cluster report this week including all new matches and ICW matches since 2019 with 20 cMs or greater. In most cases, the matches in the original clusters have remained the same, i.e. are still clustering with the same matches. The new heatmap shows the two [C8] matches are now clustering with a [C29] and a [C30] match, on the edge of the larger [C29] cluster and correlating with a cluster made up of [C54] matches.

To give a clearer picture of the clusters, here is a screenshot of my E.D.’s heatmap. It was generated using the data of her top 333 matches with 50 cMs or higher. All of the Barron-Dempster matches (highlighted in green) over 50 cMs are found in this heatmap of clusters 4 through 8.

Screenshot of part of a report generated by Jonathan Brecher’s Shared Clustering Tool
  • Clusters 4 & 5 have descendants of Mary M. DEMPSEY, d/o Seaton
  • Cluster 6 has descendants of William S., George W., Martha Ann, and Julia DEMPSEY, all children of Seaton
  • Cluster 7 has a descendant of Geneva DEMPSEY, d/o Seaton
  • Cluster 8 has only Barron-Dempster descendants
  • The Barron-Dempster matches correlate only with clusters 4 through 8. They don’t correlate with clusters 1-3 or 9-33 (not seen in this close-up of the heatmap). The correlation can be seen by the red outside of the cluster boxes.
  • Of the 35 matches shown above, 6 are mystery matches, 8 are Barron-Dempster matches, and the rest are descendants of Seaton Y. DEMPSEY and Clementine M. GOWING through six of their eight children. The two missing children are sons who served in the Civil War, died during or soon after the war, never married, and had no known descendants. The mystery matches, like the Barron-Dempster matches, correlate only with clusters 4 through 8 and are likely descendants of Seaton and Clementine through one of their children.
What Are the Odds?

I used the What Are the Odds? tool on DNA Painter to chart Mollie’s family tree down to her descendants who are matches. This is not the real purpose of the tool.

What Are the Odds? by DNA Painter

The matches, descendants of Mollie, are shaded green. I used my E.D.’s shared cMs amounts for all matches. The numbers in parenthesis are the range of cMs shared between the match and the other tests I have access to. The bottom row represents the line that I share with my cousins and is used for comparison: my great-great-grandmother Mary M. DEMPSEY, my great-grandmother Laura Belle INGRAM, my grandfather Fred R. DEMPSEY and his brother Earl S. DEMPSEY, my father’s generation represented by E.D. (1C1R), and my generation (with my cousins and brother).

What Are the Odds? by DNA Painter

The WATO tool is used to check the probability that the amount of cM shared corresponds to the relationship in the tree. As I had already used it to chart the tree of the Barron-Dempster matches, I tried doing the reverse of what is intended with the tool. I used it to determine if the amount of cM shared by E.D. with the matches would place her in the correct position in our family tree.

  • Hypothesis 2: E.D. is the child of Hypothesis 1 and grandchild of Laura Belle INGRAM scored 9 (About 3 times more likely than the next hypothesis
    This is the most likely hypothesis.)
  • Hypothesis 3: E.D. is the child of Hypothesis 2 and grandchild of Hypothesis 1 scored 3 (About 3 times more likely than the next hypothesis)
  • Hypothesis 1: E.D. is the child of Laura Belle INGRAM and grandchild of Mary M. DEMPSEY scored 1 (Possible but not significantly more likely than the other hypotheses.)

Hypothesis 2 with a score of 9 is the most likely and puts E.D. in the right place in our family tree and shows that it is possible that Mollie was the grandchild of Seaton and Clementine.

How does Mollie fit into my family tree?

Genetic genealogy uses DNA testing along with traditional genealogy. Using all of the tools mentioned above as well as genealogy research, I have come to a conclusion on how Mollie fits into my family tree.

The cluster heatmap above shows the Barron-Dempster matches are relatives of my 1C1R E.D. and share the same ancestry as the DEMPSEY-GOWING matches. The same is true for the other tests I used in this example: my brother, Danny, Laura, Sheila, and myself. The WATO tool also backs up this assumption.

If the matches who descend from Mollie Lee DEMPSTER fit into the DEMPSEY-GOWING family group, could Wesley G. DEMPSTER be an alias for a son or nephew of Seaton Y. DEMPSEY and Clementine M. GOWING?

I don’t think the relationship was a nephew as:

  1. Seaton’s brother Wilson M. DEMPSEY was found in the 1840 census with two persons in his household: himself and his wife. No children from the marriage that took place in 1839 and no children born before this marriage.
  2. Seaton’s brother Isham Coleman DEMPSEY married in 1827 in Rockbridge County, Virginia, and removed to Ross County, Ohio, by 1830. He emigrated from Ohio to Missouri in 1854.
  3. Seaton’s brother Wesley G. DEMPSEY was likely with Seaton in 1830, wasn’t found in 1840, was single in 1850, married in 1856, and died in 1890. “W. G. Dempsey left surviving him no children nor the descendants of a child, no father, no mother, no brother, no sister” per a chancery case.
  4. Seaton’s sisters Louisa J. (md. 1840) and Eliza (md. 1843) were 18 or younger and it is not likely that one of them was the mother.
  5. As the clusters are pointing to the GOWING-CRISP branch of the DEMPSEY-GOWING family group, the matches are likely related through the GOWING side, i.e. other possibilities are the two sisters of Clementine GOWING.
  6. Clementine’s sister Emmeline GOWING married William Dison LAWHORNE in 1828 and in 1840 the only male child in their household has been identified and cannot be Wesley.
  7. Clementine’s sister Martha C. “Martissa” GOWING married Wyatt F. LILLY in 1833 and in 1840 the three male children have been identified and none can be Wesley.

I believe from about 1880 George W. DEMPSEY, the only living son of Seaton Y. DEMPSEY and Clementine M. GOWING, used the alias Wesley G. DEMPSTER, and was the father of Mollie.

Consequently, Mollie Lee DEMPSTER would have been a half-sibling to George’s three children. Her descendants would share on average the same amount of DNA as the descendants of all of Seaton and Clementine’s other children. The amount shared with any of George’s descendants would not be greater as the common ancestral couple would be Seaton and Clementine. Early on in my analysis, I had not considered this and thought George’s descendants should have higher amounts of DNA which is not the case.

What else can I do to solve this mystery?

I haven’t exhausted the DNA tools to prove the possibility of Wesley G. DEMPSTER’s being the same person as George W. DEMPSEY. I’m just at a standstill as none of the Barron-Dempster matches are on any of the sites with chromosome browsers. Being able to compare the DNA segments would help to confirm I am on the right track or not.

I’ve sent messages to all of the matches. First, a short teaser asking if they were interested in figuring out who Mollie’s father was. Then messages to the same persons with the link to my second post in this series. I even mentioned the offer to upload their raw DNA file to MyHeritage and get FREE access to all DNA features. I’ve received no replies to date and none of the tests are showing up on MyHeritage. I’d hoped my messages were read even though no replies have been received.

I was only given access to E.D.’s AncestryDNA test last week. Maybe once I begin working more with her match list I will begin to make connections with people who are interested in solving the mystery.

Have I completely confused you? Have I piqued your interest in some of the tools I’m using for DNA analysis? Do you have a similar DNA mystery you are trying to solve?

© 2021, copyright Cathy Meder-Dempsey. All rights reserved.

Returning to Blogging in the New Year – Refreshed and Excited

The definition of taking a break is interrupting one’s activity briefly. When I went into hiatus the end of October I didn’t expect it to be over two months before I would come back to blogging.

I was touched by the people who reached out to me while I was missing in action. Several messaged me directly to find out if all was well. From my young 3C1R Luella who I’ve known nearly two decades to my #1 reader/commenter/blogger Amy to my follower from Brazil whose ancestors lived in the same village as my ancestors.

All were worried. They didn’t know I’d fallen into a rabbit hole, spinning down winding double helix strands carrying our DNA. It took me a while to gain my orientation and find the even more twisted ladder out of the hole.

Who’s Fault Was It?


Blaine T. Bettinger shared my post How DNA Results Helped Discover Luxembourg Emigrants in the Facebook group Genetic Genealogy Tips & Techniques.

Great blog post about how the DNA Match Labeling extension for Chrome helped solve a genealogical mystery! Genetic networks and clustering tools are the future of DNA evidence!

I had no idea I was even on Blaine’s radar and it explained a spike in traffic on my blog during the week following the post. Being noticed by Blaine was fantastic.

Even more incredible was the help I received from a member of the Genetic Genealogy Tips & Techniques group.  Jonathan Brecher sent a message offering to run a Shared Clustering tool he has developed on the AncestryDNA test I manage to help me tickle out the maternal matches.

Shared Clustering

As mentioned in the above post [over two months ago] maternal matches are few and far between as that side of my family tree is Luxembourgish with a few branches which reach into France and Germany during the periods of time when the area belonged to Luxembourg.

Jonathan’s tool is not yet available to the public. He sent a CSV file with the heat map of my matches and a list of the clusters in text format. He paid special attention to my starred matches as these were the ones I had already been able to identify as maternal.

The heat map generated 66 clusters. Four of these are for maternal matches while 61 are for paternal. One cluster remains unknown at this time but looks more paternal than maternal.

The number of matches in each cluster varies greatly. There are a dozen clusters with only 2-10 matches, 33 between 11-100, 11 between 101-200, 4 between 201-400, 5 between 401-500, and one with 705!

I pinned down the fourth maternal cluster this past week – when I was supposed to be working on this post. I felt the pull of that rabbit hole, again, and checked each match and their trees until I found the connection. They descend from immigrants, two BAUSTERT brothers who were great-grandsons of my 5th great-grandparents Matthias SCHRAMEN and Anna Barbara LEIBRICH (BURG) of Ferschweiler, Eifelkreis Bitburg-Prüm, Rheinland-Pfalz, Germany. Did the Baustert brothers know their 1C1R Nicholas SCHRAMEN had emigrated about 20 years earlier and originally settled in Iowa where they were also found?

What I’ve Been Working On

As the CSV file Jonathan sent included my notes, the paternal clusters were easily identified as coming from one of the four paternal grandparents’ branches. Some could even be associated with specific branches of a grandparent’s ancestral line.

I’ve been amending my notes on AncestryDNA to reflect the cluster number as well as a surname and possible generation. The cluster numbers are only for reference and make it easier to sort them on the AncestryDNA page using the Chrome extension AncestryDNA Helper atDNA Helper or in Genome Mate Pro when sorting the MRCA (most recent common ancestors) notes. [Note: The name of the Chrome extension was changed in April 2019 after they were notified the name was a violation of Ancestry’s trademark name.]

Cluster 40 with 13 matches is labeled GROELINGER-MERGEN(6) as the MRCA have been identified as my 4th great-grandparents (6 generations back) Johann GROELINGER and Anna Maria Benedikta MERGEN. Six of the 13 matches have been identified as descendants of this couple. I’ve sent messages and am waiting for replies.

Once the notes have been fixed on AncestryDNA, I move all matches for a cluster over to Genome Mate Pro (GMP) using another Chrome extension, Pedigree Thief (collects the match information, notes, and the shared matches). When the matches are in GMP, I begin adding the matches’ trees once again with the Pedigree Thief which reads the pedigree view of the tree and converts it to an Ahnentafel chart. GMP has a very steep learning curve and I’m still trying to assimilate and grasp the abilities of the program.

I’ve developed a routine and am slowly getting matches which have been associated with a cluster entered into GMP. Nearly half of the clusters, the smallest, have been added. The larger clusters remain to be done and I’ll be spacing them out a bit. And of course, as new matches are found on Gedmatch Genesis, FTDNA, and MyHeritage they are also added to Genome Mate Pro.

I still feel the pull of the rabbit hole but I won’t let it get in the way of my returning to a regular blogging schedule.

© 2019, copyright Cathy Meder-Dempsey. All rights reserved.