1

I am not sure about this plot if it looks good as the dots appear around the line. So I just want to ask if it looks okay. My data is quite large (almost 3000 observations) and maybe that is why it looks like this.

enter image description here

Nick Cox
  • 56,404
  • 8
  • 127
  • 185
user3288202
  • 1,325
  • 6
  • 17
  • 27
  • 5
    No, it looks fine. When you say 'why it looks like this', which particular features are you concerned about? What is it you think it should look like? – Glen_b Feb 25 '14 at 11:36
  • Thanks Glen_b. You reply relieves me. I thought it might look bad because everytime I see a good-looking residual plot, the dot would go around the line and not clustered like this, so I am not sure if mine is okay. Maybe because the size of data in the examples is less than mine. – user3288202 Feb 25 '14 at 11:48
  • Also, consider plotting standardized residual (in z-score) rather than in their original unit in the y-axis. That way you can see how many of them got beyond +/- 3 or +/- 4 SDs. And I agree that it looks fine. – Penguin_Knight Feb 25 '14 at 13:22
  • 1
    Are you talking about the 'clumping' of the fitted values at low values (below 10000, say)? That's really not an issue. – Glen_b Feb 25 '14 at 13:25
  • 2
    I wish my residual plots looked as good as this. – Hong Ooi Feb 25 '14 at 14:14
  • Never give up Hong Ooi. I usually find people here are willing to help. – user3288202 Feb 25 '14 at 14:49
  • 1
    I'd add to @Penguin_Knight one more - to plot studentized residuals either. – ttnphns Feb 25 '14 at 15:29

2 Answers2

2

Have you tried plotting this using ggplot2 in R? It has a nice semi transparency feature with the Cairo package which makes guesstimating the mean for such residual plots easy. For example you could have each point semi transparent and you could visually check if they are centered around 0. But overall by looking at the image you posted, no reason to think otherwise.

broccoli
  • 1,048
1

The plot of the residual looks good. To be very sure get the mean it should be very close to zero. In fact it should approximate to zero when you approximate to 1 decimal place

frisbee
  • 399