Is it okay to have values for dummies?

[San K]

Hi,
This is not a SPSS question but I’m sure lot of econometricians and
modeling experts are in this group who could help me out.

I am modeling demand for a utility company. This is a MIXED model
(panel data) with five years of data. Monthly dummies are used to take
care of the seasonal effects.

Unfortunately meter reads are done any day in a month. That creates
problem when I want to assign zero or one to the monthly dummy.

This is my question:
Say a property is read on the 20th of April and Jul 10th.


Is it okay to create monthly dummies as follows:
:
:
Mar = 0/31 =0
Apr = 20/30 =0.67
May = 31/31 =1
Jun = 30/30 =1
Jul = 10/31 = 0.32
Aug = 0/31 =0
:
:


Instead of:
:
:
Mar = 0
Apr = 1
May = 1
Jun = 1
Jul = 0
Aug = 1
:
:


I always thought that dummies take either zero or one and also the
number of dummies is one less than the levels!

Any help would be appreciated.

Thanks,
San

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Maguin, Eugene]

SanK,

I am not an economist.

I think you have at least two problems. One problem is that meter readings
can be taken on any day of the month so that the interval between meter
readings is variable. the second problem you have is month effects as some
months have 30 days, others 31 and sometimes February has 29 or 28. I'd also
expect weekday-weekend effects as I'd bet that consumption is not constant
over the days of the week. The third problem is seasonality effects. And
there may be more calendar effects as well.

It sounds like you know the reading date and the reading. If you discard the
first reading but keep the date of the first reading you can compute the
interval between readings and then compute the average daily use for the
month in which the meter was read. I'm assuming that while a meter may be
read on any day of the month, from month to month the same meter is read on
approximately the same day--give or take a day or two. So you will never
have a situation where the September, 2010 reading for a specific meter is
September 28 and the October reading is October 4. But you can check this in
your data.

So then I'd think your level 1 model would be

MeterReading = B0 + MonthDummy*B1 + YearDummy*B2 + .... + e

Gene Maguin



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of San
K
Sent: Friday, October 08, 2010 12:51 AM
To: [hidden email]
Subject: Is it okay to have values for dummies?

Hi,
This is not a SPSS question but I'm sure lot of econometricians and
modeling experts are in this group who could help me out.

I am modeling demand for a utility company. This is a MIXED model
(panel data) with five years of data. Monthly dummies are used to take
care of the seasonal effects.

Unfortunately meter reads are done any day in a month. That creates
problem when I want to assign zero or one to the monthly dummy.

This is my question:
Say a property is read on the 20th of April and Jul 10th.


Is it okay to create monthly dummies as follows:
:
:
Mar = 0/31 =0
Apr = 20/30 =0.67
May = 31/31 =1
Jun = 30/30 =1
Jul = 10/31 = 0.32
Aug = 0/31 =0
:
:


Instead of:
:
:
Mar = 0
Apr = 1
May = 1
Jun = 1
Jul = 0
Aug = 1
:
:


I always thought that dummies take either zero or one and also the
number of dummies is one less than the levels!

Any help would be appreciated.

Thanks,
San

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[San K]

Thank you Gene for pointing out issues with 'meter readings'. They are
all handled one way the other.

I'm still not sure if I can use various values for a dummy variable.
For example, if the reading is taken between 20th of April and 20th of
Jun then is it okay have my monthly dummies as such:
MarDummy = 0.
AprDummy = (30days-20th)/30days =0.333
MayDummy = 1
JunDummy = (20th)/30days = 0.666



On Sat, Oct 9, 2010 at 12:29 AM, Gene Maguin <[hidden email]> wrote:

> SanK,
>
> I am not an economist.
>
> I think you have at least two problems. One problem is that meter readings
> can be taken on any day of the month so that the interval between meter
> readings is variable. the second problem you have is month effects as some
> months have 30 days, others 31 and sometimes February has 29 or 28. I'd also
> expect weekday-weekend effects as I'd bet that consumption is not constant
> over the days of the week. The third problem is seasonality effects. And
> there may be more calendar effects as well.
>
> It sounds like you know the reading date and the reading. If you discard the
> first reading but keep the date of the first reading you can compute the
> interval between readings and then compute the average daily use for the
> month in which the meter was read. I'm assuming that while a meter may be
> read on any day of the month, from month to month the same meter is read on
> approximately the same day--give or take a day or two. So you will never
> have a situation where the September, 2010 reading for a specific meter is
> September 28 and the October reading is October 4. But you can check this in
> your data.
>
> So then I'd think your level 1 model would be
>
> MeterReading = B0 + MonthDummy*B1 + YearDummy*B2 + .... + e
>
> Gene Maguin
>
>
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of San
> K
> Sent: Friday, October 08, 2010 12:51 AM
> To: [hidden email]
> Subject: Is it okay to have values for dummies?
>
> Hi,
> This is not a SPSS question but I'm sure lot of econometricians and
> modeling experts are in this group who could help me out.
>
> I am modeling demand for a utility company. This is a MIXED model
> (panel data) with five years of data. Monthly dummies are used to take
> care of the seasonal effects.
>
> Unfortunately meter reads are done any day in a month. That creates
> problem when I want to assign zero or one to the monthly dummy.
>
> This is my question:
> Say a property is read on the 20th of April and Jul 10th.
>
>
> Is it okay to create monthly dummies as follows:
> :
> :
> Mar = 0/31 =0
> Apr = 20/30 =0.67
> May = 31/31 =1
> Jun = 30/30 =1
> Jul = 10/31 = 0.32
> Aug = 0/31 =0
> :
> :
>
>
> Instead of:
> :
> :
> Mar = 0
> Apr = 1
> May = 1
> Jun = 1
> Jul = 0
> Aug = 1
> :
> :
>
>
> I always thought that dummies take either zero or one and also the
> number of dummies is one less than the levels!
>
> Any help would be appreciated.
>
> Thanks,
> San
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD