Simple regression

[chamarique]

my research question is: "Can we say that the number of hours of Internet use increases with Internet purchase?" with the variables: - Use of Internet per week (Use in hours) - If respondent has already made a purchase on the Internet (1 = yes, 2 = no)

Therefore, this statement told me that my dependant variable is number of hours of Internet use and my independant variable is Online purchase. But that seems me a little weird. How can I check if I don't have to reverse my IV and my DV ? (I use SPSS) Thank you

[Ryan]

Homework assignment?

On Sat, May 7, 2016 at 1:33 PM, chamarique <[hidden email]> wrote:

my research question is: "Can we say that the number of hours of Internet use
increases with Internet purchase?" with the variables: - Use of Internet per
week (Use in hours) - If respondent has already made a purchase on the
Internet (1 = yes, 2 = no)

Therefore, this statement told me that my dependant variable is number of
hours of Internet use and my independant variable is Online purchase. But
that seems me a little weird. How can I check if I don't have to reverse my
IV and my DV ? (I use SPSS) Thank you



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Simple-regression-tp5732112.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[chamarique]

yes

[Bruce Weaver]

You asked, "How can I check if I don't have to reverse my IV and my DV?"

To answer that, you need to think about the timing of events and measurements.  For example, in one common study design in epidemiology (the cohort study), one starts with a group of people (a cohort) who are all free of the outcome of interest (e.g., an internet purchase) but whose exposure status (hours/week of internet in your example) is known.  Subjects are then followed forward over time, and (new) occurrences of the outcome are recorded.  If this was your study design, it would be clear that Purchase is dependent and hours/week explanatory.  

But this is not your study design.  It sounds to me like you have historical information for both variables.  This makes it more like a cross-sectional design (sometimes called snapshot study), where both variables are measured at the same time.  In that case, how you assign variables to explanatory and dependent roles is less clear, and usually driven by the wording of the research question.  

The wording you used in stating the research question ("Can we say that the number of hours of Internet use increases with Internet purchase?") seems to me consistent with treating hours/week as dependent on purchase.  Is that the wording that was given to you in the homework assignment?  Or have you paraphrased it?  

HTH.

p.s. - The simple linear regression you proposed is equivalent to the unequal variances version of the independent groups t-test.  See Jerry Dallal's note, for example:  

   http://www.jerrydallal.com/lhsp/treg.htm

chamarique wrote

my research question is: "Can we say that the number of hours of Internet use increases with Internet purchase?" with the variables: - Use of Internet per week (Use in hours) - If respondent has already made a purchase on the Internet (1 = yes, 2 = no)

Therefore, this statement told me that my dependant variable is number of hours of Internet use and my independant variable is Online purchase. But that seems me a little weird. How can I check if I don't have to reverse my IV and my DV ? (I use SPSS) Thank you

[Ryan]

Bruce,

You meant to say the "equal variances version" in your post script, not "unequal variances version," right?

Ryan

On Sat, May 7, 2016 at 4:38 PM, Bruce Weaver <[hidden email]> wrote:

You asked, "How can I check if I don't have to reverse my IV and my DV?"

To answer that, you need to think about the timing of events and
measurements.  For example, in one common study design in epidemiology (the
cohort study), one starts with a group of people (a cohort) who are all free
of the outcome of interest (e.g., an internet purchase) but whose exposure
status (hours/week of internet in your example) is known.  Subjects are then
followed forward over time, and (new) occurrences of the outcome are
recorded.  If this was your study design, it would be clear that Purchase is
dependent and hours/week explanatory.

But this is not your study design.  It sounds to me like you have historical
information for both variables.  This makes it more like a cross-sectional
design (sometimes called snapshot study), where both variables are measured
at the same time.  In that case, how you assign variables to explanatory and
dependent roles is less clear, and usually driven by the wording of the
research question.

The wording you used in stating the research question ("Can we say that the
number of hours of Internet use increases with Internet purchase?") seems to
me consistent with treating hours/week as dependent on purchase.  Is that
the wording that was given to you in the homework assignment?  Or have you
paraphrased it?

HTH.

p.s. - The simple linear regression you proposed is equivalent to the
unequal variances version of the independent groups t-test.  See Jerry
Dallal's note, for example:

   http://www.jerrydallal.com/lhsp/treg.htm



chamarique wrote
> my research question is: "Can we say that the number of hours of Internet
> use increases with Internet purchase?" with the variables: - Use of
> Internet per week (Use in hours) - If respondent has already made a
> purchase on the Internet (1 = yes, 2 = no)
>
> Therefore, this statement told me that my dependant variable is number of
> hours of Internet use and my independant variable is Online purchase. But
> that seems me a little weird. How can I check if I don't have to reverse
> my IV and my DV ? (I use SPSS) Thank you





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Simple-regression-tp5732112p5732116.html

Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[Rich Ulrich]

What you say does not make clear the "operational definition" (as they say)
of your two measures.  However,  "No Internet use at all" might strongly imply
"No Internet purchases." If that is the case, I think you need to find some Hypothesis
more interesting than one that has to be true.

For a test, it does not matter which you take as the IV and DV.  The Pearson r
has the same test as the regression in either direction.

If I wanted to say something useful and interesting about those data, I think
I would separate the Hours into "Zero"  and something like deciles; and plot
the increase in the fraction versus Hours. 
with some purchase.

--
Rich Ulrich

> Date: Sat, 7 May 2016 10:33:08 -0700

> From: [hidden email]
> Subject: Simple regression
> To: [hidden email]
>
> my research question is: "Can we say that the number of hours of Internet use
> increases with Internet purchase?" with the variables: - Use of Internet per
> week (Use in hours) - If respondent has already made a purchase on the
> Internet (1 = yes, 2 = no)
>
> Therefore, this statement told me that my dependant variable is number of
> hours of Internet use and my independant variable is Online purchase. But
> that seems me a little weird. How can I check if I don't have to reverse my
> IV and my DV ? (I use SPSS) Thank you
>
>

===================== 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

[chamarique]

My research question was in French, I just translated into English. This is the complete research question:
 "Can we say that the number of hours of Internet use increases with Internet purchase? This influence of Internet purchase on the number of use can it be explained by an attitude becoming more favorable as we know Internet? "
with the variables:
- Use of Internet per week (Use in hours)
- Internet attitude (Attitude_Internet 1 = very unfavorable, 7 = very favorable)
- If respondent has already made a purchase on the Internet (1 = yes, 2 = no)

For the second part of the research question I made a mediation and I change my variable purchase in no = 0 and yes = 1 because when it was 1=yes 2=no that was confusing me. My résuflats are:
I did a mediation with process in SPSS
(Variable z = mediator = Attitude_Internet
Variable x = independent variable = Purchase_Internet
Variable y = dependent variable = Use_Internet)
The results are:
a path= 1,38 (significant)
b path = 1,6517 (significant)
c path= 3,357 (significant)
c' path =  1,087 (p>0,05 -> non significant)

[Ryan]

This could lead to a variety of interesting avenues to explore if this were not a canned homework assignment. If this is a simple introductory stats course, then chances are you've been asked to perform "simple linear regression" in which purchase of the internet is treated as the independent variable (IV) and hours of use is treated as the (DV). The professor probably wants the student to be able to address some of the following:

1. Is there a significant association between the two variables?

2. What are the intercept and slope values, and how should they be interpreted?

3. How much of the variance in the DV is explained by the IV.

I also see that the professor asked the student to evaluate the potential effect of a "mediator."

I'm generally not a fan of engaging in HW assignment questions on this forum because it has the potential for taking away the learning process from the student.

Having said that, I do enjoy the comments such as those provided by Rich and Bruce as they force the student to expand his/her knowledge about equivalent statistical tests that fall under the umbrella of general linear models (e.g., independent samples t-test), and what one might consider if faced with this actual question in various real-world scenario study designs.

Ryan

On Sun, May 8, 2016 at 1:06 AM, Rich Ulrich <[hidden email]> wrote:

What you say does not make clear the "operational definition" (as they say)
of your two measures.  However,  "No Internet use at all" might strongly imply
"No Internet purchases." If that is the case, I think you need to find some Hypothesis
more interesting than one that has to be true.

For a test, it does not matter which you take as the IV and DV.  The Pearson r
has the same test as the regression in either direction.

If I wanted to say something useful and interesting about those data, I think
I would separate the Hours into "Zero"  and something like deciles; and plot
the increase in the fraction versus Hours. 
with some purchase.

--
Rich Ulrich

> Date: Sat, 7 May 2016 10:33:08 -0700

> From: [hidden email]
> Subject: Simple regression
> To: [hidden email]
>
> my research question is: "Can we say that the number of hours of Internet use
> increases with Internet purchase?" with the variables: - Use of Internet per
> week (Use in hours) - If respondent has already made a purchase on the
> Internet (1 = yes, 2 = no)
>
> Therefore, this statement told me that my dependant variable is number of
> hours of Internet use and my independant variable is Online purchase. But
> that seems me a little weird. How can I check if I don't have to reverse my
> IV and my DV ? (I use SPSS) Thank you
>
>

===================== 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

[Bruce Weaver]

Oops!  Yes, I definitely meant to say the *equal* variances version of the independent t-test is equivalent to OLS simple linear regression performed on the same data.  Thanks, Ryan.  

Cheers,
Bruce

Ryan Black wrote

Bruce,

You meant to say the "equal variances version" in your post script, not
"unequal variances version," right?

Ryan


>
> -----
> --
> Bruce Weaver
> [hidden email]
> http://sites.google.com/a/lakeheadu.ca/bweaver/
>
> "When all else fails, RTFM."
>
> NOTE: My Hotmail account is not monitored regularly.
> To send me an e-mail, please use the address shown above.
>
> --
> View this message in context:
> http://spssx-discussion.1045642.n5.nabble.com/Simple-regression-tp5732112p5732116.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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

[Bruce Weaver]

I agree that it is usually best not to get too engaged in discussing homework assignments in a forum like this, because it can veer off in directions that were never intended by the professor, and end up hurting the student more than helping.  Students who are having problems with a homework assignment should approach their professor or teaching assistant before posting here.  They know the whole context of the course, and what they are hoping students will learn from doing the assignment.  Members of a forum like this can only guess at those things.

Here's another way to look at it.  If you are teaching a course, and give a homework assignment, do you want students with questions to approach you, or to post to an online forum where they are likely to receive feedback that is not appropriate (in a variety of ways) to the context of your course?  I suspect you want the former!  

Ryan Black wrote

This could lead to a variety of interesting avenues to explore if this were
not a canned homework assignment. If this is a simple introductory stats
course, then chances are you've been asked to perform "simple linear
regression" in which purchase of the internet is treated as the independent
variable (IV) and hours of use is treated as the (DV). The professor
probably wants the student to be able to address some of the following:

1. Is there a significant association between the two variables?
2. What are the intercept and slope values, and how should they be
interpreted?
3. How much of the variance in the DV is explained by the IV.

I also see that the professor asked the student to evaluate the potential
effect of a "mediator."

I'm generally not a fan of engaging in HW assignment questions on this
forum because it has the potential for taking away the learning process
from the student.

Having said that, I do enjoy the comments such as those provided by Rich
and Bruce as they force the student to expand his/her knowledge about
equivalent statistical tests that fall under the umbrella of general linear
models (e.g., independent samples t-test), and what one might consider if
faced with this actual question in various real-world scenario study
designs.

Ryan