Why do "printcp"and "plotcp" disagree？

[A-Pai]

library(modeldata) data(credit_data) fit <- rpart(Status~.,data = credit_data) printcp(fit) plotcp(fit)

Why do "printcp"and "plotcp" disagree？

[bethatkinson]

The plot and the values of xerror in the table look like they agree to me - what doesn't look right to you?

---

From: lg21c @.> Sent: Monday, October 4, 2021 10:30 PM To: bethatkinson/rpart @.> Cc: Subscribed @.***> Subject: [EXTERNAL] [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32)

library(modeldata) data(credit_data) fit <- rpart(Status~.,data = credit_data) printcp(fit) plotcp(fit)

Why do they disagree？ [image]https://user-images.githubusercontent.com/5112397/135955177-dbe581c1-3344-4808-9f4e-bfe02cfac880.png [image]https://user-images.githubusercontent.com/5112397/135955274-ea70918b-819a-42f4-991c-097a2d6aabdf.png

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/bethatkinson/rpart/issues/32, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG53GSBHYV4LV7D32FPLUFJWPDANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[A-Pai]

My point is the cp value in the plotcp does not agree with that of printcp

The plot and the values of xerror in the table look like they agree to me - what doesn't look right to you? … ________________________________ From: lg21c @.> Sent: Monday, October 4, 2021 10:30 PM To: bethatkinson/rpart @.> Cc: Subscribed @.***> Subject: [EXTERNAL] [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32) library(modeldata) data(credit_data) fit <- rpart(Status~.,data = credit_data) printcp(fit) plotcp(fit) Why do they disagree？ [image]https://user-images.githubusercontent.com/5112397/135955177-dbe581c1-3344-4808-9f4e-bfe02cfac880.png [image]https://user-images.githubusercontent.com/5112397/135955274-ea70918b-819a-42f4-991c-097a2d6aabdf.png — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#32>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG53GSBHYV4LV7D32FPLUFJWPDANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

My point is the cp value in the plotcp does not agree with that of printcp

[bethatkinson]

Looks to be a point between - which is also accurate. Any CP value between the 2 will work. Found this line in the code (function contributed by Brian Ripley).

cp <- sqrt(cp0 * c(Inf, cp0[-length(cp0)]))

---

From: lg21c @.> Sent: Tuesday, October 5, 2021 8:01 AM To: bethatkinson/rpart @.> Cc: Atkinson, Elizabeth J. (Beth), M.S. @.>; Comment @.> Subject: [EXTERNAL] Re: [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32)

My point is the cp value in the plotcp does not agree with that of printcp

The plot and the values of xerror in the table look like they agree to me - what doesn't look right to you? … ________________________________ From: lg21c @.> Sent: Monday, October 4, 2021 10:30 PM To: bethatkinson/rpart @.> Cc: Subscribed @.***> Subject: [EXTERNAL] [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32https://github.com/bethatkinson/rpart/issues/32) library(modeldata) data(credit_data) fit <- rpart(Status~.,data = credit_data) printcp(fit) plotcp(fit) Why do they disagree？ [image]https://user-images.githubusercontent.com/5112397/135955177-dbe581c1-3344-4808-9f4e-bfe02cfac880.png [image]https://user-images.githubusercontent.com/5112397/135955274-ea70918b-819a-42f4-991c-097a2d6aabdf.png — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#32https://github.com/bethatkinson/rpart/issues/32>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG53GSBHYV4LV7D32FPLUFJWPDANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

My point is the cp value in the plotcp does not agree with that of printcp

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/bethatkinson/rpart/issues/32#issuecomment-934390170, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG564GPR3NO4EFNTCB5LUFLZJRANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[A-Pai]

Any CP value between the 2 will work

---yes，but where does the cp value in the plotcp come from？0.066，0.037，0.015，0.011

[bethatkinson]

see the code that I sent

---

From: lg21c @.> Sent: Tuesday, October 5, 2021 9:13 AM To: bethatkinson/rpart @.> Cc: Atkinson, Elizabeth J. (Beth), M.S. @.>; Comment @.> Subject: [EXTERNAL] Re: [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32)

Any CP value between the 2 will work

---yes，but where does the cp value in the plotcp come from？0.066，0.037，0.015，0.011

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/bethatkinson/rpart/issues/32#issuecomment-934452144, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG5YOP5LKASFAKKEB3OTUFMBXJANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[bgreenwell]

@A-Pai If you take a look at the rpart/R/plotcp.R file in this repo, you'll see on line 17 that the CP values are monotonically transformed using the sqrt after multiplying by the first CP value (cp0):

cp <- sqrt(cp0 * c(Inf, cp0[-length(cp0)]))

You can verify the first couple of x-axis values, for example:

> sqrt(0.064593 * 0.067783)  # sqrt(CP1 * CP0)
[1] 0.06616878
> sqrt(0.020999 * 0.067783)  # sqrt(CP2 * CP0)
[1] 0.03772765

Not sure of the exact reason (perhaps for plotting purposes), but maybe a small note could be added to the plotcp() docs to clarify.

[bethatkinson]

yep - good point.

---

From: Brandon Greenwell @.> Sent: Tuesday, October 5, 2021 9:32 PM To: bethatkinson/rpart @.> Cc: Atkinson, Elizabeth J. (Beth), M.S. @.>; Comment @.> Subject: [EXTERNAL] Re: [bethatkinson/rpart] Why do "printcp"and "plotcp" disagree？ (#32)

If you take a look at the rpart/R/plotcp.Rhttps://github.com/bethatkinson/rpart/blob/master/R/plotcp.R file in this repo, you'll see on line 17 that the CP values are monotonically transformed using the sqrt after multiplying by the first CP value:

cp <- sqrt(cp0 * c(Inf, cp0[-length(cp0)]))

Not sure of the exact reason (maybe for plotting purposes), but maybe a small note could be added to the plotcp() docs to clarify.

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/bethatkinson/rpart/issues/32#issuecomment-935288193, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACWQG54RABSAGWJ6P4QZ6MTUFOYKTANCNFSM5FKYTFFQ. Triage notifications on the go with GitHub Mobile for iOShttps://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Androidhttps://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[A-Pai]

I got it，thank you