{
  "id": 21702,
  "title": "Initial Data Analysis & Graphical Interpretation (SAS)",
  "url": "/competitions/facebook-v-predicting-check-ins/discussion/21702",
  "author_name": "",
  "post_date": "2016-06-15T13:28:00.703Z",
  "votes": 1,
  "comment_count": 2,
  "views": 526,
  "content": "<p>I just joined this competition &amp; thought of posting my very initial analysis on this data using SAS,SQL &amp; Excel. I see there are impressive amount of investigation already done by my fellow candidates .So lot info in my notes maybe redundant.Please check out the attached file &amp; let me know if this makes sense</p>",
  "messages": [
    {
      "id": "124090",
      "postDate": "06/15/2016 13:28:00",
      "content": "<p>I just joined this competition &amp; thought of posting my very initial analysis on this data using SAS,SQL &amp; Excel. I see there are impressive amount of investigation already done by my fellow candidates .So lot info in my notes maybe redundant.Please check out the attached file &amp; let me know if this makes sense</p>",
      "rawMarkdown": "I just joined this competition & thought of posting my very initial analysis on this data using SAS,SQL & Excel. I see there are impressive amount of investigation already done by my fellow candidates .So lot info in my notes maybe redundant.Please check out the attached file & let me know if this makes sense",
      "votes": null
    },
    {
      "id": "124097",
      "postDate": "06/15/2016 14:13:38",
      "content": "<p>Please let me know if you guys have any questions. Will be happy to answer !!</p>",
      "rawMarkdown": "Please let me know if you guys have any questions. Will be happy to answer !!",
      "votes": null
    },
    {
      "id": "124126",
      "postDate": "06/15/2016 18:01:45",
      "content": "<p>Here are few of the highlights about the variables &amp; model fitting. Please check out the attached document , graphs &amp; SAS code for more details</p>\n\n<p><strong>VARIABLE EXPLORATION</strong></p>\n\n<p><strong>Accuracy</strong></p>\n\n<p>This graph in my previous attachment is based out of whole dataset. It is very clear that the max occurrences of accuracy happen between 11 and 165. Beyond 225 it goes flat &amp; it shows a long tail</p>\n\n<p><strong>Time</strong></p>\n\n<p>Two big dips. Reasons could be \ni)  Can be seasonal \nii) Maybe post mid night number of  check in are less(say 2am to 6 am less check-in)</p>\n\n<p><strong>PlaceID</strong></p>\n\n<p>Check out the skewness &amp; Kurtosis from my attached Analysis . For a standard uniform distribution skew = 0 , kurtosis = 9/5=1.8. Based on the skewness &amp; Kurtosis the distributions should be nearly uniform. Also the graph attached proves it</p>\n\n<p><strong>X &amp;Y</strong> </p>\n\n<p>This looks interesting. Both X &amp; Y seems to have uniform distribution except two spikes in the beginning and end of the graph. What might be the reason??</p>\n\n<p><strong>CORRELATION AMONG VARIABLES</strong></p>\n\n<p><em>Accuracy vs.  place_id</em></p>\n\n<p>We can see maximum check in happening between accuracy scale 0 to 220. We clearly see three bands of different check in density.</p>\n\n<p><em>Correlation of Place_ID with accuracy and time</em></p>\n\n<p>Both accuracy and time are correlated. Place_id is positively correlated with accuracy where as negatively correlated with time. Based on the p value (&lt;0.001) it seems like time is very important variable. It is one the major factor based on what user decides whether to check in or not which kind of logically makes sense</p>\n\n<p><strong>MODEL FITTING</strong></p>\n\n<p>The distribution pattern of PlaceID which looks like a hike in the front and then flatten out with a long tail. Let&#8217;s see what kind of distribution it might be.As it is time based so we can try Poisson, Binomial   &amp; Cauchy model to see the fitting (Check out the attached document for implementation details).</p>\n\n<p>From the result it seems Poisson doesn't fit well. Binomial &amp; Cauchy fits better</p>",
      "rawMarkdown": "Here are few of the highlights about the variables & model fitting. Please check out the attached document , graphs & SAS code for more details\r\n\r\n**VARIABLE EXPLORATION**\r\n\r\n**Accuracy**\r\n\r\nThis graph in my previous attachment is based out of whole dataset. It is very clear that the max occurrences of accuracy happen between 11 and 165. Beyond 225 it goes flat & it shows a long tail\r\n\r\n**Time**\r\n\r\nTwo big dips. Reasons could be \r\ni)\tCan be seasonal \r\nii)\tMaybe post mid night number of  check in are less(say 2am to 6 am less check-in)\r\n\r\n\r\n**PlaceID**\r\n\r\nCheck out the skewness & Kurtosis from my attached Analysis . For a standard uniform distribution skew = 0 , kurtosis = 9/5=1.8. Based on the skewness & Kurtosis the distributions should be nearly uniform. Also the graph attached proves it\r\n\r\n\r\n**X &Y** \r\n\r\nThis looks interesting. Both X & Y seems to have uniform distribution except two spikes in the beginning and end of the graph. What might be the reason??\r\n\r\n**CORRELATION AMONG VARIABLES**\r\n\r\n*Accuracy vs.  place_id*\r\n\r\nWe can see maximum check in happening between accuracy scale 0 to 220. We clearly see three bands of different check in density.\r\n\r\n*Correlation of Place_ID with accuracy and time*\r\n\r\nBoth accuracy and time are correlated. Place_id is positively correlated with accuracy where as negatively correlated with time. Based on the p value (<0.001) it seems like time is very important variable. It is one the major factor based on what user decides whether to check in or not which kind of logically makes sense\r\n\r\n\r\n\r\n**MODEL FITTING**\r\n\r\nThe distribution pattern of PlaceID which looks like a hike in the front and then flatten out with a long tail. Let’s see what kind of distribution it might be.As it is time based so we can try Poisson, Binomial   & Cauchy model to see the fitting (Check out the attached document for implementation details).\r\n\r\nFrom the result it seems Poisson doesn't fit well. Binomial & Cauchy fits better",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 124097,
      "author_name": "abanishmishra",
      "author_url": "",
      "post_date": "06/15/2016 14:13:38",
      "content": "<p>Please let me know if you guys have any questions. Will be happy to answer !!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 124126,
      "author_name": "abanishmishra",
      "author_url": "",
      "post_date": "06/15/2016 18:01:45",
      "content": "<p>Here are few of the highlights about the variables &amp; model fitting. Please check out the attached document , graphs &amp; SAS code for more details</p>\n\n<p><strong>VARIABLE EXPLORATION</strong></p>\n\n<p><strong>Accuracy</strong></p>\n\n<p>This graph in my previous attachment is based out of whole dataset. It is very clear that the max occurrences of accuracy happen between 11 and 165. Beyond 225 it goes flat &amp; it shows a long tail</p>\n\n<p><strong>Time</strong></p>\n\n<p>Two big dips. Reasons could be \ni)  Can be seasonal \nii) Maybe post mid night number of  check in are less(say 2am to 6 am less check-in)</p>\n\n<p><strong>PlaceID</strong></p>\n\n<p>Check out the skewness &amp; Kurtosis from my attached Analysis . For a standard uniform distribution skew = 0 , kurtosis = 9/5=1.8. Based on the skewness &amp; Kurtosis the distributions should be nearly uniform. Also the graph attached proves it</p>\n\n<p><strong>X &amp;Y</strong> </p>\n\n<p>This looks interesting. Both X &amp; Y seems to have uniform distribution except two spikes in the beginning and end of the graph. What might be the reason??</p>\n\n<p><strong>CORRELATION AMONG VARIABLES</strong></p>\n\n<p><em>Accuracy vs.  place_id</em></p>\n\n<p>We can see maximum check in happening between accuracy scale 0 to 220. We clearly see three bands of different check in density.</p>\n\n<p><em>Correlation of Place_ID with accuracy and time</em></p>\n\n<p>Both accuracy and time are correlated. Place_id is positively correlated with accuracy where as negatively correlated with time. Based on the p value (&lt;0.001) it seems like time is very important variable. It is one the major factor based on what user decides whether to check in or not which kind of logically makes sense</p>\n\n<p><strong>MODEL FITTING</strong></p>\n\n<p>The distribution pattern of PlaceID which looks like a hike in the front and then flatten out with a long tail. Let&#8217;s see what kind of distribution it might be.As it is time based so we can try Poisson, Binomial   &amp; Cauchy model to see the fitting (Check out the attached document for implementation details).</p>\n\n<p>From the result it seems Poisson doesn't fit well. Binomial &amp; Cauchy fits better</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "124090": "I just joined this competition & thought of posting my very initial analysis on this data using SAS,SQL & Excel. I see there are impressive amount of investigation already done by my fellow candidates .So lot info in my notes maybe redundant.Please check out the attached file & let me know if this makes sense",
    "124097": "Please let me know if you guys have any questions. Will be happy to answer !!",
    "124126": "Here are few of the highlights about the variables & model fitting. Please check out the attached document , graphs & SAS code for more details\r\n\r\n**VARIABLE EXPLORATION**\r\n\r\n**Accuracy**\r\n\r\nThis graph in my previous attachment is based out of whole dataset. It is very clear that the max occurrences of accuracy happen between 11 and 165. Beyond 225 it goes flat & it shows a long tail\r\n\r\n**Time**\r\n\r\nTwo big dips. Reasons could be \r\ni)\tCan be seasonal \r\nii)\tMaybe post mid night number of  check in are less(say 2am to 6 am less check-in)\r\n\r\n\r\n**PlaceID**\r\n\r\nCheck out the skewness & Kurtosis from my attached Analysis . For a standard uniform distribution skew = 0 , kurtosis = 9/5=1.8. Based on the skewness & Kurtosis the distributions should be nearly uniform. Also the graph attached proves it\r\n\r\n\r\n**X &Y** \r\n\r\nThis looks interesting. Both X & Y seems to have uniform distribution except two spikes in the beginning and end of the graph. What might be the reason??\r\n\r\n**CORRELATION AMONG VARIABLES**\r\n\r\n*Accuracy vs.  place_id*\r\n\r\nWe can see maximum check in happening between accuracy scale 0 to 220. We clearly see three bands of different check in density.\r\n\r\n*Correlation of Place_ID with accuracy and time*\r\n\r\nBoth accuracy and time are correlated. Place_id is positively correlated with accuracy where as negatively correlated with time. Based on the p value (<0.001) it seems like time is very important variable. It is one the major factor based on what user decides whether to check in or not which kind of logically makes sense\r\n\r\n\r\n\r\n**MODEL FITTING**\r\n\r\nThe distribution pattern of PlaceID which looks like a hike in the front and then flatten out with a long tail. Let’s see what kind of distribution it might be.As it is time based so we can try Poisson, Binomial   & Cauchy model to see the fitting (Check out the attached document for implementation details).\r\n\r\nFrom the result it seems Poisson doesn't fit well. Binomial & Cauchy fits better"
  },
  "source": "meta"
}